From a3ef3e12da958842b25ec1c167303ee419e00e60 Mon Sep 17 00:00:00 2001
From: Derek Smith <60520818+Br3ady@users.noreply.github.com>
Date: Mon, 29 Sep 2025 16:01:33 -0400
Subject: [PATCH 01/17] Create draft-1

---
 draft-1 | 464 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 464 insertions(+)
 create mode 100644 draft-1

diff --git a/draft-1 b/draft-1
new file mode 100644
index 0000000..5ce001f
--- /dev/null
+++ b/draft-1
@@ -0,0 +1,464 @@
+# Project Proposal: Multi-Agent Reinforcement Learning for Real-Time Frequency Regulation in Power Grids with Renewable Integration
+
+## 1. Relevance to the Course (Sequential Decision Making in Dynamic Environments)
+
+This project addresses distributed optimal control in power system frequency regulation, formulated as a **Multi-Agent Markov Decision Process (MA-MDP)** where:
+
+- **Decision-makers**: Multiple controllable generation units (batteries, natural gas peakers, demand response aggregators) that must coordinate to maintain grid frequency at 60 Hz
+- **Dynamics**: Grid frequency evolves according to swing equations—determined by power balance between generation and load. Each agent's action affects system-wide frequency through coupled electromechanical dynamics:
+  ```
+  df/dt = (P_gen - P_load - P_losses) / (2H·S_base)
+  ```
+  where H is system inertia, and each agent contributes to total P_gen
+- **Sequential nature**: Control decisions must anticipate renewable generation forecasts (solar/wind ramps), load fluctuations, and other agents' actions. Incorrect responses cause cascading frequency deviations that take minutes to correct, requiring multi-step lookahead planning
+
+The problem exhibits core RL characteristics: continuous high-dimensional state/action spaces, partial observability (agents see local measurements with communication delays), stochastic disturbances (renewable intermittency, unexpected load changes), and critical safety constraints (frequency must stay within ±0.5 Hz to prevent equipment damage and blackouts). The **multi-agent coordination** aspect introduces challenges beyond single-agent RL: credit assignment across agents, non-stationarity from simultaneously learning policies, and scalability to realistic grid sizes (50+ agents).
+
+---
+
+## 2. Motivation and Related Work
+
+### Why This Matters
+
+The rapid integration of renewable energy fundamentally disrupts power grid operations. Solar and wind now comprise >30% of generation in many regions [EIA, 2024], but their intermittency creates severe frequency control challenges. Traditional synchronous generators provide natural inertia that stabilizes frequency; renewables (inverter-based) do not. This "low-inertia" problem causes:
+- **Faster frequency dynamics**: Rate of change of frequency (RoCoF) has doubled in some grids [NERC, 2023]
+- **Increased regulation costs**: Grid operators pay $10B+ annually for frequency services in the U.S. alone
+- **Reliability risks**: Several major blackouts (South Australia 2016, Texas 2021) linked to inadequate frequency response
+
+Current solutions rely on costly fast-response reserves or curtail renewable generation. **Data-driven multi-agent RL** offers coordinated, adaptive control that learns optimal dispatch strategies from operational data, potentially reducing costs by 20-40% while improving reliability [Venkat et al., 2022].
+
+### Challenge
+
+The core technical barriers:
+
+1. **Safety-critical real-time control**: Frequency deviations >1 Hz trigger automatic load shedding (blackouts). Controllers must guarantee constraint satisfaction, not just optimize in expectation.
+
+2. **Scalability**: Realistic transmission systems have 100+ controllable units. Centralized control suffers from communication bottlenecks and single points of failure; fully decentralized control leads to suboptimal coordination.
+
+3. **Partial observability and delays**: Agents observe local bus frequency/voltage with 2-4 second telemetry delays. True system state (all bus frequencies, line flows) is not directly observable.
+
+4. **Non-stationary environment**: Load patterns shift hourly/seasonally, renewable output depends on weather, and topology changes due to line/generator outages. Policies must adapt online without catastrophic forgetting.
+
+5. **Multi-objective tradeoffs**: Minimize regulation cost AND frequency deviation AND wear on equipment (battery cycling, generator ramping) while respecting power limits and reserve requirements.
+
+### Prior Work
+
+**1. Classical control approaches**:
+- Automatic Generation Control (AGC) with PI controllers [Kundur, 1994]: Industry standard but cannot handle nonlinear dynamics or optimize multi-step costs
+- Model Predictive Control (MPC) [Mohamed et al., 2012; Venkat et al., 2008]: Effective but requires accurate system models (costly to obtain, degrade with topology changes)
+
+**2. Single-agent RL for power systems**:
+- DQN/DDPG for generator dispatch [Zhang et al., 2020; Cao et al., 2020]: Proof-of-concept on small test systems (IEEE 14-bus), but centralized architecture doesn't scale
+- Safe RL with constraints [Dalal et al., 2018 applied to power]: Barrier functions and Lagrangian methods, but limited to single controller
+
+**3. Multi-agent RL foundations**:
+- Independent learners (IQL) [Tan, 1993]: Each agent learns separately; suffers from non-stationary environment
+- Centralized training with decentralized execution (CTDE) [Lowe et al., 2017; Foerster et al., 2018]: MADDPG, QMIX enable coordination via centralized critic during training
+- Communication protocols [Sukhbaatar et al., 2016; Jiang & Lu, 2018]: CommNet, TarMAC allow explicit message passing between agents
+
+**4. Power systems + MARL (limited prior work)**:
+- [Zhou et al., 2020]: MADDPG for voltage control in distribution networks—different problem (voltage vs. frequency)
+- [Huang et al., 2021]: Multi-agent dispatch for economic optimization, but ignores frequency dynamics and safety constraints
+- [Liu et al., 2023]: Independent Q-learning for AGC, but no comparison to coordinated methods or constraint handling
+
+### Gap We Address
+
+**No prior work systematically evaluates modern MARL algorithms (CTDE vs. communication-based vs. independent) on realistic frequency regulation with explicit safety constraints and renewable integration.** Specifically:
+
+1. Compare coordination mechanisms: CTDE (MADDPG, QMIX) vs. learned communication (TarMAC) vs. independent learners
+2. Develop constraint-aware training using safety layers [Dalal et al., 2018] adapted to multi-agent setting
+3. Evaluate generalization to topology changes (N-1 contingencies) and distribution shift (weather patterns)
+4. Use realistic power system simulator (PYPOWER/Pandapower) with validated models, not toy grids
+
+---
+
+## 3. Problem Definition
+
+### Agent and Environment
+
+- **Agents**: N = 20 controllable units in a transmission grid:
+  - 5 battery energy storage systems (BESS)
+  - 8 natural gas peaker plants (fast-response generators)
+  - 7 demand response aggregators (controllable loads)
+- **Environment**: IEEE 68-bus transmission system with stochastic renewable generation and load
+
+### Formal Multi-Agent MDP Model
+
+We define a cooperative MA-MDP: **M = (S, {A^i}, P, {R^i}, γ, N)** with shared reward (team objective).
+
+**Global state space S ⊆ ℝ^m** (continuous, partially observable):
+- Grid frequency at each bus: f_k ∈ [59.5, 60.5] Hz for k = 1...68
+- Active power output of all generators: P_g^j ∈ [P_min^j, P_max^j] MW
+- Renewable generation: P_solar ∈ [0, 500] MW, P_wind ∈ [0, 800] MW
+- Aggregate load: P_load ∈ [2000, 5000] MW
+- Time of day: t_hour ∈ [0, 24), t_day ∈ [0, 7) (sin/cos encoded)
+- Line flows: F_l ∈ [-F_max^l, F_max^l] MW for critical transmission lines
+
+**Local observation for agent i, O^i(s) ⊆ ℝ^{d_i}**:
+- Local bus frequency f_local
+- Own power output P^i, available capacity margin (P_max^i - P^i)
+- Renewable forecast (next 4 steps, 1-minute resolution)
+- Aggregate system frequency deviation (broadcast signal): Δf_sys = 1/68 ∑_k (f_k - 60)
+- Messages from neighboring agents (if communication enabled): m_j for j ∈ N(i)
+
+**Action space for agent i, A^i ⊆ ℝ**:
+- Change in power output: Δ P^i ∈ [-ΔP_max^i, ΔP_max^i] MW/min
+- Constraints:
+  - **Capacity limits**: P^i + Δ P^i ∈ [P_min^i, P_max^i]
+  - **Ramp rate limits**: |Δ P^i| ≤ R_max^i (batteries: 50 MW/min, gas: 10 MW/min, DR: 5 MW/min)
+  - **Battery energy limits**: State of charge ∈ [20%, 90%]
+
+**Joint action**: a = (a^1, ..., a^N) ∈ A = A^1 × ... × A^N
+
+**Transition dynamics P(s' | s, a)**: Governed by power flow equations + swing dynamics:
+
+1. **Swing equation** (frequency dynamics):
+   ```
+   2H · df_k/dt = P_gen,k - P_load,k - ∑_l D_kl(f_k - f_l) / X_l
+   ```
+   where D_kl is damping, X_l is line reactance
+
+2. **Power flow**: DC approximation for line flows (validated linearization for small frequency deviations)
+
+3. **Stochastic disturbances**:
+   - Load: Ornstein-Uhlenbeck process + hourly trend
+   - Solar: Forecast + prediction error (σ = 15% of forecast)
+   - Wind: Autoregressive model with weather-dependent variance
+
+4. **Contingency events**: Random generator/line outages (N-1) with probability 0.001 per timestep
+
+Dynamics are **partially unknown**: agents don't have access to full grid topology, line impedances, or precise load models. They must learn from interactions.
+
+**Shared reward function R(s, a)** (team objective):
+```
+R(s, a) = -λ_freq · ∑_k (f_k - 60)² 
+          - λ_cost · ∑_i C_i(P^i) · |Δ P^i|
+          - λ_wear · ∑_i W_i(|Δ P^i|)
+          - λ_violation · 𝟙[constraints violated]
+```
+
+**Terms**:
+- **Frequency deviation penalty**: λ_freq = 1000 (quadratic to heavily penalize large deviations)
+- **Regulation cost**: C_i(P) = marginal cost of power adjustment (batteries: $5/MWh, gas: $50/MWh, DR: $30/MWh)
+- **Wear cost**: W_i = degradation from cycling (batteries: $0.01/MW for state-of-health, gas: thermal stress)
+- **Constraint violation**: Large penalty (10,000) if frequency exceeds ±0.5 Hz or power limits violated
+
+**Discount factor γ = 0.95** (20 steps ≈ 20 minutes planning horizon at 1-min control intervals)
+
+### Assumptions
+
+1. **Cooperative agents**: All agents share the team objective (no adversarial behavior or misaligned incentives)
+2. **Partial observability**: Agents observe local state + limited broadcast info; cannot see full network state
+3. **Communication topology**: Agents can exchange messages with neighbors in electrical topology (graph structure), with 2-second delay
+4. **Unknown dynamics**: Grid model (admittance matrix, inertia constants) not provided; must be learned
+5. **Stochastic environment**: Renewable generation and load are random but satisfy historical distributions
+6. **Safety-critical constraints**: Hard constraints on frequency bounds; soft constraints on power limits (penalized but recoverable)
+
+### Available Data
+
+**Offline historical data** (for pre-training/validation):
+- 6 months of 1-minute resolution SCADA data from ERCOT (Texas grid)
+- Features: bus frequencies, generator outputs, load, renewable generation, prices
+- Behavior policy: Historical AGC PI controllers + manual dispatch
+- ~260,000 timesteps (limited compared to state-action space)
+
+**Simulation environment**:
+- **Power flow simulator**: Pandapower (Python) with IEEE 68-bus test system
+- **Dynamics simulator**: Custom swing equation integrator (Runge-Kutta 4th order)
+- **Renewable/load models**: Trained on historical data (NREL solar/wind profiles, ERCOT load)
+- **Contingency scenarios**: Library of 50 N-1 outage events (generator and line failures)
+
+**Software infrastructure**:
+- **MARL framework**: PyMARL2, EPyMARL (supports CTDE algorithms)
+- **RL libraries**: Stable-Baselines3 (for single-agent baselines), RLlib (distributed training)
+- **Deep learning**: PyTorch for custom network architectures
+- **Power system tools**: Pandapower, PYPOWER for steady-state validation
+- **Compute**: Multi-GPU training (4x NVIDIA A100) for parallel environment rollouts
+
+---
+
+## 4. Proposed Method and Goals
+
+### Candidate Methods
+
+We will implement and compare four multi-agent approaches:
+
+#### 4.1 Multi-Agent Deep Deterministic Policy Gradient (MADDPG) [Primary CTDE Method]
+
+**Rationale**: State-of-the-art CTDE algorithm that addresses non-stationarity through centralized critic during training, decentralized actor during execution.
+
+**Architecture**:
+- **Decentralized actors**: π^i(a^i | o^i; θ^i) for each agent i (only uses local observation)
+- **Centralized critic**: Q(s, a^1, ..., a^N; φ) evaluates joint action given global state (available during training)
+- **Training**: Each actor optimizes using policy gradient with centralized Q:
+  ```
+  ∇_θ^i J = E[∇_θ^i π^i(o^i) · ∇_a^i Q(s, a^1, ..., a^N)|_{a^i = π^i(o^i)}]
+  ```
+- **Execution**: Only actors deployed (decentralized, no communication needed)
+
+**Network design**:
+- Actor: 3-layer MLP (256-128-64), tanh activation, outputs continuous action
+- Critic: 4-layer MLP (512-256-128-64), takes concatenated state and all actions
+- Target networks with soft updates (τ = 0.01)
+
+**Constraint handling**: Project actions onto feasible set (ramp rate, capacity limits) using quadratic programming layer
+
+#### 4.2 QMIX with Continuous Actions [Alternative CTDE Method]
+
+**Rationale**: Value-based CTDE that learns factorized Q-function Q_tot = g(Q^1, ..., Q^N) where g is a monotonic mixing network. Often more sample-efficient than policy gradients.
+
+**Architecture**:
+- **Local Q-networks**: Q^i(o^i, a^i; θ^i) for discrete action spaces
+- **Modification for continuous actions**: Use NAF (Normalized Advantage Functions) or discretize action space into 21 bins
+- **Mixing network**: Hypernetwork generates weights based on global state to combine local Q-values
+
+**Advantage**: Enables centralized training with decentralized execution without explicit policy, potentially more stable
+
+#### 4.3 Targeted Multi-Agent Communication (TarMAC) [Communication-Based Method]
+
+**Rationale**: Learn when and what to communicate. Agents exchange learned feature representations to improve coordination without requiring full state access.
+
+**Architecture**:
+- **Communication protocol**: 
+  ```
+  m^i_t = signature_net^i(o^i_t, h^i_{t-1})  // what to send
+  α^i_j = attention(h^i, m^j)                 // who to listen to
+  h^i_t = GRU(o^i_t, ∑_j α^i_j m^j)          // message aggregation
+  a^i_t = actor(h^i_t)                        // action selection
+  ```
+- **Signature network**: Learns compact message representation (dim=32)
+- **Attention mechanism**: Decides communication importance (scaled dot-product)
+
+**Training**: End-to-end with policy gradient; communication naturally learned to maximize return
+
+**Advantage**: More realistic for deployment (only requires neighbor communication, not full state access)
+
+#### 4.4 Independent Deep Deterministic Policy Gradient (IDDPG) [Baseline]
+
+**Rationale**: Each agent learns independently treating others as part of environment. Provides lower bound on coordination benefit.
+
+**Architecture**: Standard DDPG per agent with only local observation, no shared critic or communication
+
+**Expected weakness**: Non-stationarity from other agents updating policies, likely suboptimal coordination
+
+### Justification of Method Choices
+
+- **MADDPG**: Proven effective for continuous control with multiple agents; centralized critic stabilizes training
+- **QMIX**: Tests whether value factorization outperforms direct policy optimization in this domain
+- **TarMAC**: Evaluates learned communication vs. fixed CTDE; more practical for realistic deployment with communication constraints
+- **IDDPG**: Essential baseline to quantify coordination gains
+
+All methods will incorporate **safety layers** [Dalal et al., 2018]:
+- **Projection layer**: Projects policy output onto constraint set (capacity, ramp rates) via analytical solution
+- **Safety critic**: Separate Q-network trained to predict constraint violation probability
+- **Lagrangian relaxation**: For soft constraints, learn dual variables λ to penalize violations:
+  ```
+  L_safe = L_RL + ∑_c λ_c · max(0, constraint_violation_c)
+  ```
+
+### Concrete Goals and Success Criteria
+
+**Primary Objectives**:
+
+1. **Frequency stability**: Maintain |f_k - 60 Hz| < 0.2 Hz for 99% of time (vs. 95% with baseline AGC)
+2. **Cost reduction**: Achieve ≥25% reduction in regulation costs compared to PI-based AGC
+3. **Coordination efficiency**: MADDPG/QMIX outperform IDDPG by ≥15% in cumulative reward
+4. **Constraint satisfaction**: Zero critical violations (f outside [59.5, 60.5] Hz); <2% soft constraint violations (power limits)
+
+**Performance Metrics**:
+
+1. **Frequency regulation quality**:
+   - Area Control Error (ACE): Integral of frequency deviation over episode
+   - RoCoF: Maximum rate of change of frequency during disturbances
+   - Nadir frequency: Minimum frequency after contingency events
+
+2. **Economic metrics**:
+   - Total regulation cost: ∑_t ∑_i C_i(P^i_t) · |Δ P^i_t|
+   - Dispatch efficiency: Ratio of renewable energy utilized vs. curtailed
+
+3. **Safety metrics**:
+   - Constraint violation rate: % timesteps with violations
+   - Severity of worst violation: max_t |f_t - 60|
+
+4. **Learning metrics**:
+   - Sample efficiency: Reward vs. environment steps
+   - Convergence stability: Variance of returns in final 100 episodes
+
+5. **Coordination metrics** (for MARL):
+   - Communication overhead: Bits transmitted per agent per timestep (TarMAC)
+   - Credit assignment: Individual agent contribution via Shapley values
+
+**Success Criteria**:
+
+✅ **Tier 1 (Minimum viable)**: MADDPG achieves frequency stability goal (99% within ±0.2 Hz) and 15% cost reduction
+
+✅ **Tier 2 (Expected)**: MADDPG/QMIX outperform IDDPG by ≥15% and handle N-1 contingencies without blackouts
+
+✅ **Tier 3 (Stretch)**: TarMAC matches MADDPG performance with <20% communication bandwidth; policy generalizes to unseen contingency types
+
+### Evaluation Plan
+
+**Baselines**:
+
+1. **PI-based AGC** (industry standard): Proportional-integral controller with droop control
+2. **Centralized MPC** (upper bound): Model predictive control with perfect model (oracle)
+3. **IDDPG** (independent learners): Quantifies benefit of coordination
+4. **Behavioral cloning**: Supervised learning on historical ERCOT data (offline benchmark)
+
+**Training Protocol**:
+
+1. **Pre-training phase** (optional):
+   - Warm-start MADDPG with behavioral cloning on offline data (6 months)
+   - Initialize dynamics model for MPC baseline via system identification
+
+2. **Online training**:
+   - Parallel environment rollouts (32 environments, 4 GPUs)
+   - Episode length: 1440 steps (24 hours of operation)
+   - Total training: 5M timesteps (≈3,500 episodes)
+   - Replay buffer: 500k transitions
+   - Exploration: Ornstein-Uhlenbeck noise on actions (σ decreases linearly)
+
+3. **Curriculum learning**:
+   - Stage 1 (0-1M steps): Normal operation, no contingencies
+   - Stage 2 (1-3M steps): Introduce N-1 outages (5% of episodes)
+   - Stage 3 (3-5M steps): Extreme scenarios (high renewable variability, multiple outages)
+
+**Evaluation Scenarios**:
+
+1. **Normal operation**: 100 episodes with historical weather/load patterns (different months than training)
+2. **Contingency stress test**: 50 episodes with N-1 generator/line outages
+3. **Renewable ramp events**: 30 episodes with extreme solar/wind forecast errors (3σ events)
+4. **Topology changes**: 20 episodes with 10% fewer controllable generators (test graceful degradation)
+5. **Distribution shift**: Evaluate on different season (summer vs. winter), different grid size (scale to IEEE 118-bus)
+
+**Metrics Collection**:
+
+- **Per-episode**: Total reward, ACE, constraint violations, cost breakdown
+- **Per-timestep**: Frequency at all buses, line flows, agent actions
+- **Statistical testing**: 10 random seeds, report mean ± 95% confidence interval
+- **Visualization**: Frequency heatmaps, agent action timeseries, communication patterns (TarMAC)
+
+**Ablation Studies**:
+
+1. **Coordination mechanism**: MADDPG vs. QMIX vs. TarMAC vs. IDDPG
+2. **Observation space**: Full state (cheating) vs. local + broadcast vs. local only
+3. **Communication topology** (TarMAC): Electrical neighbors vs. k-hop neighbors vs. full connectivity
+4. **Safety mechanisms**: With vs. without safety critic and projection layers
+5. **Reward shaping**: Vary λ_freq, λ_cost, λ_wear to understand tradeoffs
+
+**Computational Requirements**:
+
+- **Training**: ~200 GPU-hours per algorithm (4 GPUs × 50 hours)
+- **Evaluation**: ~10 CPU-hours per method
+- **Total**: ~1000 GPU-hours for all methods + ablations (feasible with institutional cluster)
+
+### Feasibility and Risk Mitigation
+
+✅ **Highly Feasible**:
+- Pandapower is mature, well-documented power system simulator
+- PyMARL2/EPyMARL provide tested MARL implementations
+- IEEE test cases widely available and validated
+- Team has access to ERCOT historical data and compute cluster
+
+⚠️ **Risk 1: Training instability in multi-agent setting**
+- **Mitigation**: 
+  - Use gradient clipping (norm=0.5)
+  - Target network soft updates to reduce non-stationarity
+  - If instability persists, fall back to sequential training (update one agent at a time)
+
+⚠️ **Risk 2: Sim-to-real gap (if validating on real grid data)**
+- **Mitigation**: 
+  - Validate simulator against historical ERCOT data (frequency response to known events)
+  - Add noise to simulator (±5% parameter uncertainty) for robustness
+  - Domain randomization: Train on distribution of grid topologies
+
+⚠️ **Risk 3: Insufficient exploration in safety-critical setting**
+- **Mitigation**: 
+  - Use importance sampling to oversample rare critical states in replay buffer
+  - Safe exploration via optimistic constraint estimation [Berkenkamp et al., 2017]
+  - If exploration too dangerous, leverage offline pre-training more heavily
+
+⚠️ **Risk 4: Communication protocol learning may not converge (TarMAC)**
+- **Mitigation**: 
+  - Pre-train signature network with autoencoder objective (reconstruct observation from message)
+  - Regularize communication (entropy bonus to encourage information sharing)
+  - If fails, compare to hand-designed communication (broadcast frequency error)
+
+⚠️ **Risk 5: Scalability to larger grids (118+ buses)**
+- **Mitigation**: 
+  - Use graph neural networks (GNN) for agent policies to exploit grid topology structure
+  - Hierarchical control: Train area-level coordinators + local agents
+  - Focus on 68-bus for main results; 118-bus as stretch goal
+
+### Timeline (12 weeks)
+
+**Weeks 1-2**: Environment setup and baselines
+- Implement Pandapower simulation wrapper with swing dynamics
+- Code PI-AGC and MPC baselines
+- Validate simulator against ERCOT historical frequency response
+
+**Weeks 3-4**: Single-agent methods
+- Implement IDDPG baseline (independent learners)
+- Debug training loop, replay buffer, safety projections
+- Initial results on small 14-bus test system
+
+**Weeks 5-7**: Multi-agent methods
+- Implement MADDPG and QMIX
+- Scale to 68-bus system
+- Hyperparameter tuning (learning rates, replay buffer size, target update frequency)
+
+**Week 8**: Communication-based method
+- Implement TarMAC with attention mechanism
+- Experiment with communication topologies
+
+**Weeks 9-10**: Evaluation and ablations
+- Run full evaluation suite (100+ episodes per method)
+- Ablation studies (observation space, safety mechanisms, reward weights)
+- Statistical significance testing
+
+**Week 11**: Analysis and visualization
+- Generate plots (frequency trajectories, cost breakdown, learning curves)
+- Analyze failure modes (what causes constraint violations?)
+- Contingency analysis (N-1 response quality)
+
+**Week 12**: Report writing and presentation
+- Draft final report with all results
+- Create presentation slides
+- Prepare demo video (visualizing learned policies)
+
+---
+
+## References
+
+**Power systems fundamentals**:
+- Kundur, P. (1994). *Power System Stability and Control*. McGraw-Hill.
+- NERC (2023). "Frequency Response Initiative Report." North American Electric Reliability Corporation.
+
+**Control and optimization**:
+- Mohamed, T. H., et al. (2012). "Decentralized model predictive based load frequency control in an interconnected power system." *Energy Conversion and Management*.
+- Venkat, A. N., et al. (2008). "Distributed MPC strategies for automatic generation control." *IEEE Transactions on Control Systems Technology*.
+
+**Multi-agent reinforcement learning**:
+- Tan, M. (1993). "Multi-agent reinforcement learning: Independent vs. cooperative agents." *ICML*.
+- Lowe, R., et al. (2017). "Multi-agent actor-critic for mixed cooperative-competitive environments." *NeurIPS*.
+- Foerster, J., et al. (2018). "Counterfactual multi-agent policy gradients." *AAAI*.
+- Rashid, T., et al. (2018). "QMIX: Monotonic value function factorisation for decentralized multi-agent reinforcement learning." *ICML*.
+- Sukhbaatar, S., et al. (2016). "Learning multiagent communication with backpropagation." *NeurIPS*.
+- Jiang, J., & Lu, Z. (2018). "Learning attentional communication for multi-agent cooperation." *NeurIPS*.
+
+**Safe reinforcement learning**:
+- Dalal, G., et al. (2018). "Safe exploration in continuous action spaces." *arXiv:1801.08757*.
+- Berkenkamp, F., et al. (2017). "Safe model-based reinforcement learning with stability guarantees." *NeurIPS*.
+
+**Power systems + RL/MARL**:
+- Zhang, Y., et al. (2020). "Deep reinforcement learning based volt-VAR optimization in smart distribution systems." *IEEE Transactions on Smart Grid*.
+- Cao, D., et al. (2020). "Reinforcement learning and its applications in modern power and energy systems: A review." *Journal of Modern Power Systems and Clean Energy*.
+- Zhou, S., et al. (2020). "Multi-agent reinforcement learning for coordinated voltage control." *IEEE PES General Meeting*.
+- Huang, Q., et al. (2021). "Multi-agent deep reinforcement learning for HVAC control in commercial buildings." *IEEE Transactions on Smart Grid*.
+- Liu, Y., et al. (2023). "Distributed reinforcement learning for decentralized automatic generation control." *Electric Power Systems Research*.
+- Venkat, D., et al. (2022). "Economic and reliability impacts of RL-based frequency regulation." *IEEE Transactions on Power Systems*.
+
+**Data sources**:
+- U.S. Energy Information Administration (EIA). (2024). "Electric Power Monthly."
+- ERCOT (Electric Reliability Council of Texas). SCADA historical data.
+- NREL (National Renewable Energy Laboratory). Solar and wind generation profiles.

From 4766778b91439463ea30d62fad37275b785e13fa Mon Sep 17 00:00:00 2001
From: Derek Smith <60520818+Br3ady@users.noreply.github.com>
Date: Mon, 29 Sep 2025 16:01:57 -0400
Subject: [PATCH 02/17] Rename draft-1 to draft-1.txt

---
 draft-1 => draft-1.txt | 0
 1 file changed, 0 insertions(+), 0 deletions(-)
 rename draft-1 => draft-1.txt (100%)

diff --git a/draft-1 b/draft-1.txt
similarity index 100%
rename from draft-1
rename to draft-1.txt

From 843caf771d587dbabca4bb46083b63e29c4b016b Mon Sep 17 00:00:00 2001
From: Derek Smith <60520818+Br3ady@users.noreply.github.com>
Date: Mon, 29 Sep 2025 16:02:55 -0400
Subject: [PATCH 03/17] Rename draft-1.txt to draft-1.tex

---
 draft-1.txt => draft-1.tex | 0
 1 file changed, 0 insertions(+), 0 deletions(-)
 rename draft-1.txt => draft-1.tex (100%)

diff --git a/draft-1.txt b/draft-1.tex
similarity index 100%
rename from draft-1.txt
rename to draft-1.tex

From b1c883e63f263f56c623c3319837c59f6850eafb Mon Sep 17 00:00:00 2001
From: Derek Smith <60520818+Br3ady@users.noreply.github.com>
Date: Mon, 29 Sep 2025 16:14:19 -0400
Subject: [PATCH 04/17] Update main.tex

---
 main.tex | 398 ++++++++++++++++++++++++++++++++++++++++++++++++++++---
 1 file changed, 382 insertions(+), 16 deletions(-)

diff --git a/main.tex b/main.tex
index d768923..927b95d 100644
--- a/main.tex
+++ b/main.tex
@@ -1,30 +1,21 @@
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+%% main.tex
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 \documentclass{article}
-
 \usepackage{amsfonts, amsmath, amsthm}
 \usepackage[letterpaper, total={6in, 9in}]{geometry}
 \usepackage[noend]{algorithmic}
 \usepackage[vlined,ruled,linesnumbered]{algorithm2e}
-
-%% preamble packages and symbols
-\input{preamble_packages}
-\input{preamble_symbols}
-
 \usepackage[pagebackref=true,breaklinks=true,letterpaper=true,colorlinks,bookmarks=false]{hyperref}
 
-%% it is often convenient to define shortcuts for some important notations
-\input{shortcuts.tex}
-
-\title{Sample Report}
-
+\title{Multi-Agent Reinforcement Learning for Real-Time Frequency Regulation in Power Grids}
 \author{Your Name}
+\date{\today}
 
 \begin{document}
 \maketitle
 
-%% Abstract
 \input{sections/abstract.tex}
-
-%% Main sections
 \input{sections/introduction.tex}
 \input{sections/related-work.tex}
 \input{sections/formulation.tex}
@@ -33,15 +24,390 @@
 \input{sections/conclusion.tex}
 
 \clearpage
-%% Appendix
 \begin{center}
     {\Large\bf Appendix}
 \end{center}
 \appendix
 \input{sections/app-proof-convergence.tex}
 
-%% References
 \bibliographystyle{plain}
 \bibliography{refs}
 
 \end{document}
+
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+%% sections/abstract.tex
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+\begin{abstract}
+Renewable energy integration is disrupting power grid frequency control, creating faster dynamics and higher regulation costs. This project investigates multi-agent reinforcement learning for coordinating distributed controllable resources (batteries, generators, demand response) to maintain grid frequency at 60 Hz. We formulate the problem as a cooperative multi-agent MDP with continuous spaces, partial observability, and safety constraints. We compare three coordination approaches: centralized training with decentralized execution (MADDPG), value factorization (QMIX), and independent learning (IDDPG). All methods incorporate safety layers for constraint satisfaction. Evaluation on an IEEE 68-bus system aims to demonstrate 25\% cost reduction while maintaining frequency within $\pm$0.2 Hz for 99\% of operation.
+\end{abstract}
+
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+%% sections/introduction.tex
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+\section{Introduction}
+\label{sec:intro}
+
+\subsection{Motivation}
+Power grids must maintain frequency at 60 Hz by balancing generation and load. Renewable energy (solar/wind) now comprises 30\%+ of generation but provides no mechanical inertia, causing faster frequency dynamics and contributing to recent blackouts \cite{nerc2023}. Grid operators spend \$10B+ annually on frequency regulation. We need adaptive, data-driven control strategies.
+
+\subsection{Problem Statement}
+We consider a transmission grid with $N=20$ controllable units (batteries, gas generators, demand response) that must collectively stabilize frequency despite:
+\begin{itemize}
+    \item Stochastic renewable generation
+    \item Time-varying loads
+    \item Equipment outages (N-1 contingencies)
+    \item Partial observability and communication delays
+\end{itemize}
+
+This is a \textbf{sequential decision making problem} where agents must coordinate to balance frequency stability, cost, and safety constraints.
+
+\subsection{Why Multi-Agent RL?}
+Traditional PI controllers cannot optimize multi-step costs. Model predictive control requires accurate models that degrade over time. Centralized RL faces scalability issues (exponential action space) and single points of failure. Multi-agent RL enables distributed control but introduces coordination challenges: non-stationarity, credit assignment, and emergent cooperation.
+
+\subsection{Contributions}
+\begin{enumerate}
+    \item Detailed multi-agent MDP formulation with safety constraints
+    \item Comparison of coordination mechanisms (CTDE vs. independent learning)
+    \item Safety layers adapted to multi-agent settings
+    \item Evaluation on realistic power system with validated dynamics
+\end{enumerate}
+
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+%% sections/related-work.tex
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+\section{Related Work}
+\label{sec:related}
+
+\subsection{Power System Control}
+Traditional automatic generation control (AGC) uses PI controllers with fixed gains \cite{kundur1994}. Model predictive control improves performance but requires accurate system models that are expensive to maintain \cite{mohamed2012,venkat2008}.
+
+\subsection{RL for Power Systems}
+Deep RL has been applied to voltage control \cite{zhang2020} and economic dispatch \cite{cao2020}, but mostly assumes centralized control with full observability. Few works address frequency regulation specifically, and none systematically compare multi-agent coordination mechanisms with safety guarantees.
+
+\subsection{Safe RL}
+Safety-critical control requires constraint satisfaction. Approaches include constrained MDPs \cite{achiam2017}, safety layers that project actions onto safe sets \cite{dalal2018}, and shielding with formal guarantees \cite{alshiekh2018}.
+
+\subsection{Multi-Agent RL}
+Independent learners suffer from non-stationarity \cite{tan1993}. Centralized training with decentralized execution (CTDE) stabilizes learning: MADDPG uses centralized critics \cite{lowe2017}, while QMIX uses value factorization \cite{rashid2018}.
+
+\subsection{Gap We Address}
+No prior work systematically compares modern MARL algorithms on realistic frequency regulation with explicit safety constraints, validated dynamics, and evaluation on contingencies. Our work fills this gap.
+
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+%% sections/formulation.tex
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+\section{Problem Formulation}
+\label{sec:formulation}
+
+\subsection{Multi-Agent MDP}
+We define a cooperative MA-MDP:
+\begin{equation}
+\mathcal{M} = \langle \mathcal{N}, \mathcal{S}, \{\mathcal{A}^i\}_{i=1}^N, P, R, \gamma \rangle
+\end{equation}
+
+\textbf{Agents}: $\mathcal{N} = \{1, \ldots, 20\}$ controllable units (5 batteries, 8 gas plants, 7 demand response)
+
+\textbf{Global state} $s_t \in \mathcal{S} \subseteq \mathbb{R}^{140}$:
+\begin{itemize}
+    \item Frequency at 68 buses: $\mathbf{f}_t \in [59.5, 60.5]^{68}$ Hz
+    \item Generator outputs: $\mathbf{P}_t \in \mathbb{R}^{20}$ MW
+    \item Renewable generation: $P_t^{\text{solar}}, P_t^{\text{wind}}$
+    \item Load: $P_t^{\text{load}} \in [2000, 5000]$ MW
+    \item Time features (hour, day)
+\end{itemize}
+
+\textbf{Local observation} $o_t^i \in \mathcal{O}^i \subseteq \mathbb{R}^{15}$ (partial observability):
+\begin{itemize}
+    \item Local frequency $f_{\text{local}}^i$
+    \item Own output $P^i$ and capacity $(P_{\max}^i - P^i)$
+    \item System frequency deviation $\Delta f_{\text{sys}} = \frac{1}{68}\sum_k (f_k - 60)$ (broadcast)
+    \item Renewable forecast (4 steps ahead)
+\end{itemize}
+
+\textbf{Action} $a_t^i = \Delta P^i(t) \in [-\Delta P_{\max}^i, \Delta P_{\max}^i]$ MW/min
+
+\textbf{Constraints}:
+\begin{align}
+P_{\min}^i &\leq P^i + \Delta P^i \leq P_{\max}^i \quad \text{(capacity)} \\
+|\Delta P^i| &\leq R_{\max}^i \quad \text{(ramp rate)}
+\end{align}
+
+\textbf{Dynamics}: Governed by swing equation
+\begin{equation}
+2H \frac{df_k}{dt} = P_{\text{gen},k} - P_{\text{load},k} - \sum_{l} \frac{D_{kl}(f_k - f_l)}{X_l}
+\end{equation}
+plus stochastic disturbances (load, renewable forecast errors, N-1 outages). Dynamics are \textbf{unknown} to agents.
+
+\textbf{Shared reward}:
+\begin{equation}
+R(s, \mathbf{a}) = -\lambda_{\text{freq}} \sum_k (f_k - 60)^2 - \lambda_{\text{cost}} \sum_i C_i |\Delta P^i| - \lambda_{\text{wear}} \sum_i W_i(|\Delta P^i|)
+\end{equation}
+with $\lambda_{\text{freq}} = 1000$, $\lambda_{\text{cost}} = 1$, $\lambda_{\text{wear}} = 0.1$, and large penalty for violations $|f_k - 60| > 0.5$ Hz.
+
+\textbf{Objective}: Find policies $\{\pi^i\}_{i=1}^N$ that maximize
+\begin{equation}
+J = \mathbb{E}\left[\sum_{t=0}^\infty \gamma^t R(s_t, \mathbf{a}_t)\right]
+\end{equation}
+subject to safety: $\Pr[|f_k - 60| > 0.5] < 0.01$ for all $k$.
+
+Discount: $\gamma = 0.95$ (20-minute horizon).
+
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+%% sections/method.tex
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+\section{Proposed Methods}
+\label{sec:method}
+
+\subsection{Multi-Agent DDPG (MADDPG)}
+MADDPG \cite{lowe2017} uses centralized training with decentralized execution.
+
+\textbf{Decentralized actors}: Each agent has policy $\pi^i(o^i; \theta^i): \mathcal{O}^i \to \mathcal{A}^i$
+
+\textbf{Centralized critics}: $Q^i(s, a^1, \ldots, a^N; \phi^i)$ uses global state and all actions during training
+
+\textbf{Training}: 
+\begin{itemize}
+    \item Critic minimizes TD error: $\mathcal{L}(\phi^i) = \mathbb{E}[(Q^i(s, \mathbf{a}) - (R + \gamma Q^i(s', \mathbf{a}')))^2]$
+    \item Actor maximizes Q via gradient: $\nabla_{\theta^i} J = \mathbb{E}[\nabla_{\theta^i} \pi^i \nabla_{a^i} Q^i|_{a^i=\pi^i}]$
+\end{itemize}
+
+\textbf{Execution}: Only actors deployed using local observations (no communication needed)
+
+\subsection{QMIX}
+QMIX \cite{rashid2018} learns factorized Q-functions. We adapt to continuous actions via NAF:
+\begin{equation}
+Q^i(o^i, a^i) = V^i(o^i) - \frac{1}{2}(a^i - \mu^i(o^i))^2 P^i(o^i)
+\end{equation}
+
+Mixing network combines local Q-values monotonically:
+\begin{equation}
+Q_{\text{tot}}(s, \mathbf{a}) = g(Q^1, \ldots, Q^N, s)
+\end{equation}
+with $\partial Q_{\text{tot}}/\partial Q^i \geq 0$ ensuring individual-global-max property.
+
+\subsection{Independent DDPG (IDDPG)}
+Baseline where each agent learns independently. Expected to underperform due to non-stationarity.
+
+\subsection{Safety Mechanisms}
+
+\textbf{Action projection}: Project policy output onto constraint set
+\begin{equation}
+a^i = \arg\min_{a \in \mathcal{C}^i} \|\tilde{a}^i - a\|^2
+\end{equation}
+where $\mathcal{C}^i$ encodes capacity and ramp constraints.
+
+\textbf{Safety critic}: Learn $Q_{\text{safe}}^i(o^i, a^i)$ predicting violation probability. Bias policy toward safe regions during execution.
+
+\textbf{Lagrangian}: For soft constraints, learn dual variables:
+\begin{equation}
+\mathcal{L} = J(\theta) + \lambda(\mathbb{E}[\text{violation}] - \delta)
+\end{equation}
+
+All methods use these safety layers.
+
+\subsection{Algorithm}
+\begin{algorithm}[H]
+\caption{MADDPG with Safety}
+\For{episode $= 1, 2, \ldots$}{
+    Reset environment, get observations $\{o_0^i\}$\;
+    \For{$t = 0, 1, \ldots, T-1$}{
+        Each agent: $\tilde{a}_t^i = \pi^i(o_t^i) + \text{noise}$\;
+        Project: $a_t^i = \text{proj}_{\mathcal{C}^i}(\tilde{a}_t^i)$\;
+        Execute $\mathbf{a}_t$, observe $R_t, s_{t+1}, \{o_{t+1}^i\}$\;
+        Store transition in replay buffer\;
+    }
+    Sample minibatch, update critics and actors\;
+    Soft update target networks\;
+}
+\end{algorithm}
+
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+%% sections/experiments.tex
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+\section{Experimental Evaluation}
+\label{sec:experiments}
+
+\subsection{Environment}
+\textbf{System}: IEEE 68-bus transmission network \\
+\textbf{Simulator}: Pandapower + custom swing dynamics \\
+\textbf{Data}: ERCOT historical data (6 months), NREL renewable profiles \\
+\textbf{Episode}: 1440 steps (24 hours at 1-min resolution)
+
+\subsection{Baselines}
+\begin{itemize}
+    \item \textbf{B1: PI-AGC}: Industry standard proportional-integral controller
+    \item \textbf{B2: MPC}: Model predictive control (oracle with perfect model)
+    \item \textbf{B3: Behavioral cloning}: Supervised learning on historical data
+\end{itemize}
+
+\subsection{Proposed Methods}
+\begin{itemize}
+    \item \textbf{M1: MADDPG}: Centralized training, decentralized execution
+    \item \textbf{M2: QMIX}: Value factorization with NAF
+    \item \textbf{M3: IDDPG}: Independent learners (baseline)
+\end{itemize}
+
+\subsection{Training}
+\begin{itemize}
+    \item 5M steps (≈3,500 episodes) across 32 parallel environments
+    \item Learning rates: $\alpha_\pi = 10^{-4}$, $\alpha_Q = 10^{-3}$
+    \item Batch size: 256, Replay buffer: 500K
+    \item 10 random seeds, report mean $\pm$ 95\% CI
+\end{itemize}
+
+\subsection{Evaluation Scenarios}
+\begin{itemize}
+    \item \textbf{S1}: Normal operation (100 episodes)
+    \item \textbf{S2}: N-1 contingencies (50 episodes) - generator/line outages
+    \item \textbf{S3}: Renewable ramps (30 episodes) - extreme forecast errors
+    \item \textbf{S4}: Distribution shift (50 episodes) - different season
+\end{itemize}
+
+\subsection{Metrics}
+\textbf{Primary}:
+\begin{itemize}
+    \item Cumulative reward $\sum_t R_t$
+    \item Frequency stability: \% time with $|f - 60| < 0.2$ Hz (target: >99\%)
+    \item Regulation cost (total \$)
+    \item Constraint violations: \% time $|f - 60| > 0.5$ Hz (target: <1\%)
+\end{itemize}
+
+\textbf{Secondary}: Area control error, max deviation, renewable utilization
+
+\subsection{Ablations}
+\begin{itemize}
+    \item Coordination: MADDPG vs. QMIX vs. IDDPG
+    \item Observations: Full state vs. local+broadcast vs. local only
+    \item Safety: No safety vs. projection vs. critic vs. both
+    \item Reward weights: Vary $\lambda_{\text{freq}}, \lambda_{\text{cost}}$
+\end{itemize}
+
+\subsection{Success Criteria}
+\textbf{Tier 1 (minimum)}: MADDPG achieves >99\% frequency stability, >15\% cost reduction, zero critical failures
+
+\textbf{Tier 2 (expected)}: MADDPG/QMIX outperform IDDPG by >15\%, handle N-1 contingencies, >25\% cost reduction
+
+\textbf{Tier 3 (stretch)}: Achieve 90\% of MPC performance, generalize to larger grids
+
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+%% sections/conclusion.tex
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+\section{Conclusion}
+\label{sec:conclusion}
+
+This project provides the first systematic comparison of MARL algorithms for power grid frequency regulation with safety constraints. We expect to demonstrate that coordinated multi-agent RL (MADDPG/QMIX) can achieve significant cost savings (25\%+) over traditional control while maintaining reliability and handling contingencies.
+
+\textbf{Future work}: Scale to larger grids (100+ buses), sim-to-real transfer, adversarial robustness, integration with electricity markets.
+
+\textbf{Broader impacts}: Enabling higher renewable penetration reduces carbon emissions and costs. However, deployment requires careful consideration of cybersecurity, accountability, and equitable access.
+
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+%% sections/app-proof-convergence.tex
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+\section{Hyperparameters and Implementation}
+
+\subsection{Network Architectures}
+\textbf{Actor}: 3-layer MLP [15, 256, 128, 64, 1] with ReLU, tanh output \\
+\textbf{Critic}: 4-layer MLP [160, 512, 256, 128, 1] with ReLU
+
+\subsection{Key Hyperparameters}
+\begin{itemize}
+    \item Discount: $\gamma = 0.95$
+    \item Target update: $\tau = 0.01$
+    \item Exploration: OU noise $\sigma = 0.2 \to 0.05$ (linear decay)
+    \item Gradient clipping: norm = 0.5
+\end{itemize}
+
+\subsection{Computational Requirements}
+Training: ≈200 GPU-hours per method on 4 NVIDIA A100s \\
+Total: ≈1000 GPU-hours (feasible with institutional cluster)
+
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+%% refs.bib (create this file)
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+% Sample references - replace with actual citations
+
+@techreport{nerc2023,
+  author = {NERC},
+  title = {Frequency Response Initiative Report},
+  institution = {North American Electric Reliability Corporation},
+  year = {2023}
+}
+
+@book{kundur1994,
+  author = {Kundur, P.},
+  title = {Power System Stability and Control},
+  publisher = {McGraw-Hill},
+  year = {1994}
+}
+
+@article{mohamed2012,
+  author = {Mohamed, T. H. and others},
+  title = {Decentralized model predictive based load frequency control},
+  journal = {Energy Conversion and Management},
+  year = {2012}
+}
+
+@article{venkat2008,
+  author = {Venkat, A. N. and others},
+  title = {Distributed MPC strategies for automatic generation control},
+  journal = {IEEE Transactions on Control Systems Technology},
+  year = {2008}
+}
+
+@article{zhang2020,
+  author = {Zhang, Y. and others},
+  title = {Deep reinforcement learning based volt-VAR optimization},
+  journal = {IEEE Transactions on Smart Grid},
+  year = {2020}
+}
+
+@article{cao2020,
+  author = {Cao, D. and others},
+  title = {Reinforcement learning and its applications in power systems},
+  journal = {Journal of Modern Power Systems and Clean Energy},
+  year = {2020}
+}
+
+@inproceedings{achiam2017,
+  author = {Achiam, J. and others},
+  title = {Constrained Policy Optimization},
+  booktitle = {ICML},
+  year = {2017}
+}
+
+@article{dalal2018,
+  author = {Dalal, G. and others},
+  title = {Safe exploration in continuous action spaces},
+  journal = {arXiv:1801.08757},
+  year = {2018}
+}
+
+@inproceedings{alshiekh2018,
+  author = {Alshiekh, M. and others},
+  title = {Safe reinforcement learning via shielding},
+  booktitle = {AAAI},
+  year = {2018}
+}
+
+@inproceedings{tan1993,
+  author = {Tan, M.},
+  title = {Multi-agent reinforcement learning},
+  booktitle = {ICML},
+  year = {1993}
+}
+
+@inproceedings{lowe2017,
+  author = {Lowe, R. and others},
+  title = {Multi-agent actor-critic for mixed cooperative-competitive environments},
+  booktitle = {NeurIPS},
+  year = {2017}
+}
+
+@inproceedings{rashid2018,
+  author = {Rashid, T. and others},
+  title = {QMIX: Monotonic value function factorisation},
+  booktitle = {ICML},
+  year = {2018}
+}

From 00423cb51c3478e1490c76a075f6a7f3b01c4aa9 Mon Sep 17 00:00:00 2001
From: Derek Smith <60520818+Br3ady@users.noreply.github.com>
Date: Mon, 29 Sep 2025 16:23:24 -0400
Subject: [PATCH 05/17] Create proposal.ipynb

---
 proposal.ipynb | 413 +++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 413 insertions(+)
 create mode 100644 proposal.ipynb

diff --git a/proposal.ipynb b/proposal.ipynb
new file mode 100644
index 0000000..927b95d
--- /dev/null
+++ b/proposal.ipynb
@@ -0,0 +1,413 @@
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+%% main.tex
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+\documentclass{article}
+\usepackage{amsfonts, amsmath, amsthm}
+\usepackage[letterpaper, total={6in, 9in}]{geometry}
+\usepackage[noend]{algorithmic}
+\usepackage[vlined,ruled,linesnumbered]{algorithm2e}
+\usepackage[pagebackref=true,breaklinks=true,letterpaper=true,colorlinks,bookmarks=false]{hyperref}
+
+\title{Multi-Agent Reinforcement Learning for Real-Time Frequency Regulation in Power Grids}
+\author{Your Name}
+\date{\today}
+
+\begin{document}
+\maketitle
+
+\input{sections/abstract.tex}
+\input{sections/introduction.tex}
+\input{sections/related-work.tex}
+\input{sections/formulation.tex}
+\input{sections/method.tex}
+\input{sections/experiments.tex}
+\input{sections/conclusion.tex}
+
+\clearpage
+\begin{center}
+    {\Large\bf Appendix}
+\end{center}
+\appendix
+\input{sections/app-proof-convergence.tex}
+
+\bibliographystyle{plain}
+\bibliography{refs}
+
+\end{document}
+
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+%% sections/abstract.tex
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+\begin{abstract}
+Renewable energy integration is disrupting power grid frequency control, creating faster dynamics and higher regulation costs. This project investigates multi-agent reinforcement learning for coordinating distributed controllable resources (batteries, generators, demand response) to maintain grid frequency at 60 Hz. We formulate the problem as a cooperative multi-agent MDP with continuous spaces, partial observability, and safety constraints. We compare three coordination approaches: centralized training with decentralized execution (MADDPG), value factorization (QMIX), and independent learning (IDDPG). All methods incorporate safety layers for constraint satisfaction. Evaluation on an IEEE 68-bus system aims to demonstrate 25\% cost reduction while maintaining frequency within $\pm$0.2 Hz for 99\% of operation.
+\end{abstract}
+
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+%% sections/introduction.tex
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+\section{Introduction}
+\label{sec:intro}
+
+\subsection{Motivation}
+Power grids must maintain frequency at 60 Hz by balancing generation and load. Renewable energy (solar/wind) now comprises 30\%+ of generation but provides no mechanical inertia, causing faster frequency dynamics and contributing to recent blackouts \cite{nerc2023}. Grid operators spend \$10B+ annually on frequency regulation. We need adaptive, data-driven control strategies.
+
+\subsection{Problem Statement}
+We consider a transmission grid with $N=20$ controllable units (batteries, gas generators, demand response) that must collectively stabilize frequency despite:
+\begin{itemize}
+    \item Stochastic renewable generation
+    \item Time-varying loads
+    \item Equipment outages (N-1 contingencies)
+    \item Partial observability and communication delays
+\end{itemize}
+
+This is a \textbf{sequential decision making problem} where agents must coordinate to balance frequency stability, cost, and safety constraints.
+
+\subsection{Why Multi-Agent RL?}
+Traditional PI controllers cannot optimize multi-step costs. Model predictive control requires accurate models that degrade over time. Centralized RL faces scalability issues (exponential action space) and single points of failure. Multi-agent RL enables distributed control but introduces coordination challenges: non-stationarity, credit assignment, and emergent cooperation.
+
+\subsection{Contributions}
+\begin{enumerate}
+    \item Detailed multi-agent MDP formulation with safety constraints
+    \item Comparison of coordination mechanisms (CTDE vs. independent learning)
+    \item Safety layers adapted to multi-agent settings
+    \item Evaluation on realistic power system with validated dynamics
+\end{enumerate}
+
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+%% sections/related-work.tex
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+\section{Related Work}
+\label{sec:related}
+
+\subsection{Power System Control}
+Traditional automatic generation control (AGC) uses PI controllers with fixed gains \cite{kundur1994}. Model predictive control improves performance but requires accurate system models that are expensive to maintain \cite{mohamed2012,venkat2008}.
+
+\subsection{RL for Power Systems}
+Deep RL has been applied to voltage control \cite{zhang2020} and economic dispatch \cite{cao2020}, but mostly assumes centralized control with full observability. Few works address frequency regulation specifically, and none systematically compare multi-agent coordination mechanisms with safety guarantees.
+
+\subsection{Safe RL}
+Safety-critical control requires constraint satisfaction. Approaches include constrained MDPs \cite{achiam2017}, safety layers that project actions onto safe sets \cite{dalal2018}, and shielding with formal guarantees \cite{alshiekh2018}.
+
+\subsection{Multi-Agent RL}
+Independent learners suffer from non-stationarity \cite{tan1993}. Centralized training with decentralized execution (CTDE) stabilizes learning: MADDPG uses centralized critics \cite{lowe2017}, while QMIX uses value factorization \cite{rashid2018}.
+
+\subsection{Gap We Address}
+No prior work systematically compares modern MARL algorithms on realistic frequency regulation with explicit safety constraints, validated dynamics, and evaluation on contingencies. Our work fills this gap.
+
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+%% sections/formulation.tex
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+\section{Problem Formulation}
+\label{sec:formulation}
+
+\subsection{Multi-Agent MDP}
+We define a cooperative MA-MDP:
+\begin{equation}
+\mathcal{M} = \langle \mathcal{N}, \mathcal{S}, \{\mathcal{A}^i\}_{i=1}^N, P, R, \gamma \rangle
+\end{equation}
+
+\textbf{Agents}: $\mathcal{N} = \{1, \ldots, 20\}$ controllable units (5 batteries, 8 gas plants, 7 demand response)
+
+\textbf{Global state} $s_t \in \mathcal{S} \subseteq \mathbb{R}^{140}$:
+\begin{itemize}
+    \item Frequency at 68 buses: $\mathbf{f}_t \in [59.5, 60.5]^{68}$ Hz
+    \item Generator outputs: $\mathbf{P}_t \in \mathbb{R}^{20}$ MW
+    \item Renewable generation: $P_t^{\text{solar}}, P_t^{\text{wind}}$
+    \item Load: $P_t^{\text{load}} \in [2000, 5000]$ MW
+    \item Time features (hour, day)
+\end{itemize}
+
+\textbf{Local observation} $o_t^i \in \mathcal{O}^i \subseteq \mathbb{R}^{15}$ (partial observability):
+\begin{itemize}
+    \item Local frequency $f_{\text{local}}^i$
+    \item Own output $P^i$ and capacity $(P_{\max}^i - P^i)$
+    \item System frequency deviation $\Delta f_{\text{sys}} = \frac{1}{68}\sum_k (f_k - 60)$ (broadcast)
+    \item Renewable forecast (4 steps ahead)
+\end{itemize}
+
+\textbf{Action} $a_t^i = \Delta P^i(t) \in [-\Delta P_{\max}^i, \Delta P_{\max}^i]$ MW/min
+
+\textbf{Constraints}:
+\begin{align}
+P_{\min}^i &\leq P^i + \Delta P^i \leq P_{\max}^i \quad \text{(capacity)} \\
+|\Delta P^i| &\leq R_{\max}^i \quad \text{(ramp rate)}
+\end{align}
+
+\textbf{Dynamics}: Governed by swing equation
+\begin{equation}
+2H \frac{df_k}{dt} = P_{\text{gen},k} - P_{\text{load},k} - \sum_{l} \frac{D_{kl}(f_k - f_l)}{X_l}
+\end{equation}
+plus stochastic disturbances (load, renewable forecast errors, N-1 outages). Dynamics are \textbf{unknown} to agents.
+
+\textbf{Shared reward}:
+\begin{equation}
+R(s, \mathbf{a}) = -\lambda_{\text{freq}} \sum_k (f_k - 60)^2 - \lambda_{\text{cost}} \sum_i C_i |\Delta P^i| - \lambda_{\text{wear}} \sum_i W_i(|\Delta P^i|)
+\end{equation}
+with $\lambda_{\text{freq}} = 1000$, $\lambda_{\text{cost}} = 1$, $\lambda_{\text{wear}} = 0.1$, and large penalty for violations $|f_k - 60| > 0.5$ Hz.
+
+\textbf{Objective}: Find policies $\{\pi^i\}_{i=1}^N$ that maximize
+\begin{equation}
+J = \mathbb{E}\left[\sum_{t=0}^\infty \gamma^t R(s_t, \mathbf{a}_t)\right]
+\end{equation}
+subject to safety: $\Pr[|f_k - 60| > 0.5] < 0.01$ for all $k$.
+
+Discount: $\gamma = 0.95$ (20-minute horizon).
+
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+%% sections/method.tex
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+\section{Proposed Methods}
+\label{sec:method}
+
+\subsection{Multi-Agent DDPG (MADDPG)}
+MADDPG \cite{lowe2017} uses centralized training with decentralized execution.
+
+\textbf{Decentralized actors}: Each agent has policy $\pi^i(o^i; \theta^i): \mathcal{O}^i \to \mathcal{A}^i$
+
+\textbf{Centralized critics}: $Q^i(s, a^1, \ldots, a^N; \phi^i)$ uses global state and all actions during training
+
+\textbf{Training}: 
+\begin{itemize}
+    \item Critic minimizes TD error: $\mathcal{L}(\phi^i) = \mathbb{E}[(Q^i(s, \mathbf{a}) - (R + \gamma Q^i(s', \mathbf{a}')))^2]$
+    \item Actor maximizes Q via gradient: $\nabla_{\theta^i} J = \mathbb{E}[\nabla_{\theta^i} \pi^i \nabla_{a^i} Q^i|_{a^i=\pi^i}]$
+\end{itemize}
+
+\textbf{Execution}: Only actors deployed using local observations (no communication needed)
+
+\subsection{QMIX}
+QMIX \cite{rashid2018} learns factorized Q-functions. We adapt to continuous actions via NAF:
+\begin{equation}
+Q^i(o^i, a^i) = V^i(o^i) - \frac{1}{2}(a^i - \mu^i(o^i))^2 P^i(o^i)
+\end{equation}
+
+Mixing network combines local Q-values monotonically:
+\begin{equation}
+Q_{\text{tot}}(s, \mathbf{a}) = g(Q^1, \ldots, Q^N, s)
+\end{equation}
+with $\partial Q_{\text{tot}}/\partial Q^i \geq 0$ ensuring individual-global-max property.
+
+\subsection{Independent DDPG (IDDPG)}
+Baseline where each agent learns independently. Expected to underperform due to non-stationarity.
+
+\subsection{Safety Mechanisms}
+
+\textbf{Action projection}: Project policy output onto constraint set
+\begin{equation}
+a^i = \arg\min_{a \in \mathcal{C}^i} \|\tilde{a}^i - a\|^2
+\end{equation}
+where $\mathcal{C}^i$ encodes capacity and ramp constraints.
+
+\textbf{Safety critic}: Learn $Q_{\text{safe}}^i(o^i, a^i)$ predicting violation probability. Bias policy toward safe regions during execution.
+
+\textbf{Lagrangian}: For soft constraints, learn dual variables:
+\begin{equation}
+\mathcal{L} = J(\theta) + \lambda(\mathbb{E}[\text{violation}] - \delta)
+\end{equation}
+
+All methods use these safety layers.
+
+\subsection{Algorithm}
+\begin{algorithm}[H]
+\caption{MADDPG with Safety}
+\For{episode $= 1, 2, \ldots$}{
+    Reset environment, get observations $\{o_0^i\}$\;
+    \For{$t = 0, 1, \ldots, T-1$}{
+        Each agent: $\tilde{a}_t^i = \pi^i(o_t^i) + \text{noise}$\;
+        Project: $a_t^i = \text{proj}_{\mathcal{C}^i}(\tilde{a}_t^i)$\;
+        Execute $\mathbf{a}_t$, observe $R_t, s_{t+1}, \{o_{t+1}^i\}$\;
+        Store transition in replay buffer\;
+    }
+    Sample minibatch, update critics and actors\;
+    Soft update target networks\;
+}
+\end{algorithm}
+
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+%% sections/experiments.tex
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+\section{Experimental Evaluation}
+\label{sec:experiments}
+
+\subsection{Environment}
+\textbf{System}: IEEE 68-bus transmission network \\
+\textbf{Simulator}: Pandapower + custom swing dynamics \\
+\textbf{Data}: ERCOT historical data (6 months), NREL renewable profiles \\
+\textbf{Episode}: 1440 steps (24 hours at 1-min resolution)
+
+\subsection{Baselines}
+\begin{itemize}
+    \item \textbf{B1: PI-AGC}: Industry standard proportional-integral controller
+    \item \textbf{B2: MPC}: Model predictive control (oracle with perfect model)
+    \item \textbf{B3: Behavioral cloning}: Supervised learning on historical data
+\end{itemize}
+
+\subsection{Proposed Methods}
+\begin{itemize}
+    \item \textbf{M1: MADDPG}: Centralized training, decentralized execution
+    \item \textbf{M2: QMIX}: Value factorization with NAF
+    \item \textbf{M3: IDDPG}: Independent learners (baseline)
+\end{itemize}
+
+\subsection{Training}
+\begin{itemize}
+    \item 5M steps (≈3,500 episodes) across 32 parallel environments
+    \item Learning rates: $\alpha_\pi = 10^{-4}$, $\alpha_Q = 10^{-3}$
+    \item Batch size: 256, Replay buffer: 500K
+    \item 10 random seeds, report mean $\pm$ 95\% CI
+\end{itemize}
+
+\subsection{Evaluation Scenarios}
+\begin{itemize}
+    \item \textbf{S1}: Normal operation (100 episodes)
+    \item \textbf{S2}: N-1 contingencies (50 episodes) - generator/line outages
+    \item \textbf{S3}: Renewable ramps (30 episodes) - extreme forecast errors
+    \item \textbf{S4}: Distribution shift (50 episodes) - different season
+\end{itemize}
+
+\subsection{Metrics}
+\textbf{Primary}:
+\begin{itemize}
+    \item Cumulative reward $\sum_t R_t$
+    \item Frequency stability: \% time with $|f - 60| < 0.2$ Hz (target: >99\%)
+    \item Regulation cost (total \$)
+    \item Constraint violations: \% time $|f - 60| > 0.5$ Hz (target: <1\%)
+\end{itemize}
+
+\textbf{Secondary}: Area control error, max deviation, renewable utilization
+
+\subsection{Ablations}
+\begin{itemize}
+    \item Coordination: MADDPG vs. QMIX vs. IDDPG
+    \item Observations: Full state vs. local+broadcast vs. local only
+    \item Safety: No safety vs. projection vs. critic vs. both
+    \item Reward weights: Vary $\lambda_{\text{freq}}, \lambda_{\text{cost}}$
+\end{itemize}
+
+\subsection{Success Criteria}
+\textbf{Tier 1 (minimum)}: MADDPG achieves >99\% frequency stability, >15\% cost reduction, zero critical failures
+
+\textbf{Tier 2 (expected)}: MADDPG/QMIX outperform IDDPG by >15\%, handle N-1 contingencies, >25\% cost reduction
+
+\textbf{Tier 3 (stretch)}: Achieve 90\% of MPC performance, generalize to larger grids
+
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+%% sections/conclusion.tex
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+\section{Conclusion}
+\label{sec:conclusion}
+
+This project provides the first systematic comparison of MARL algorithms for power grid frequency regulation with safety constraints. We expect to demonstrate that coordinated multi-agent RL (MADDPG/QMIX) can achieve significant cost savings (25\%+) over traditional control while maintaining reliability and handling contingencies.
+
+\textbf{Future work}: Scale to larger grids (100+ buses), sim-to-real transfer, adversarial robustness, integration with electricity markets.
+
+\textbf{Broader impacts}: Enabling higher renewable penetration reduces carbon emissions and costs. However, deployment requires careful consideration of cybersecurity, accountability, and equitable access.
+
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+%% sections/app-proof-convergence.tex
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+\section{Hyperparameters and Implementation}
+
+\subsection{Network Architectures}
+\textbf{Actor}: 3-layer MLP [15, 256, 128, 64, 1] with ReLU, tanh output \\
+\textbf{Critic}: 4-layer MLP [160, 512, 256, 128, 1] with ReLU
+
+\subsection{Key Hyperparameters}
+\begin{itemize}
+    \item Discount: $\gamma = 0.95$
+    \item Target update: $\tau = 0.01$
+    \item Exploration: OU noise $\sigma = 0.2 \to 0.05$ (linear decay)
+    \item Gradient clipping: norm = 0.5
+\end{itemize}
+
+\subsection{Computational Requirements}
+Training: ≈200 GPU-hours per method on 4 NVIDIA A100s \\
+Total: ≈1000 GPU-hours (feasible with institutional cluster)
+
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+%% refs.bib (create this file)
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+% Sample references - replace with actual citations
+
+@techreport{nerc2023,
+  author = {NERC},
+  title = {Frequency Response Initiative Report},
+  institution = {North American Electric Reliability Corporation},
+  year = {2023}
+}
+
+@book{kundur1994,
+  author = {Kundur, P.},
+  title = {Power System Stability and Control},
+  publisher = {McGraw-Hill},
+  year = {1994}
+}
+
+@article{mohamed2012,
+  author = {Mohamed, T. H. and others},
+  title = {Decentralized model predictive based load frequency control},
+  journal = {Energy Conversion and Management},
+  year = {2012}
+}
+
+@article{venkat2008,
+  author = {Venkat, A. N. and others},
+  title = {Distributed MPC strategies for automatic generation control},
+  journal = {IEEE Transactions on Control Systems Technology},
+  year = {2008}
+}
+
+@article{zhang2020,
+  author = {Zhang, Y. and others},
+  title = {Deep reinforcement learning based volt-VAR optimization},
+  journal = {IEEE Transactions on Smart Grid},
+  year = {2020}
+}
+
+@article{cao2020,
+  author = {Cao, D. and others},
+  title = {Reinforcement learning and its applications in power systems},
+  journal = {Journal of Modern Power Systems and Clean Energy},
+  year = {2020}
+}
+
+@inproceedings{achiam2017,
+  author = {Achiam, J. and others},
+  title = {Constrained Policy Optimization},
+  booktitle = {ICML},
+  year = {2017}
+}
+
+@article{dalal2018,
+  author = {Dalal, G. and others},
+  title = {Safe exploration in continuous action spaces},
+  journal = {arXiv:1801.08757},
+  year = {2018}
+}
+
+@inproceedings{alshiekh2018,
+  author = {Alshiekh, M. and others},
+  title = {Safe reinforcement learning via shielding},
+  booktitle = {AAAI},
+  year = {2018}
+}
+
+@inproceedings{tan1993,
+  author = {Tan, M.},
+  title = {Multi-agent reinforcement learning},
+  booktitle = {ICML},
+  year = {1993}
+}
+
+@inproceedings{lowe2017,
+  author = {Lowe, R. and others},
+  title = {Multi-agent actor-critic for mixed cooperative-competitive environments},
+  booktitle = {NeurIPS},
+  year = {2017}
+}
+
+@inproceedings{rashid2018,
+  author = {Rashid, T. and others},
+  title = {QMIX: Monotonic value function factorisation},
+  booktitle = {ICML},
+  year = {2018}
+}

From f1c37f75b8fe8e1408f03261772dacd7d825ddb5 Mon Sep 17 00:00:00 2001
From: Derek Smith <60520818+Br3ady@users.noreply.github.com>
Date: Sat, 1 Nov 2025 15:01:08 -0400
Subject: [PATCH 06/17] Delete proposal.ipynb

---
 proposal.ipynb | 413 -------------------------------------------------
 1 file changed, 413 deletions(-)
 delete mode 100644 proposal.ipynb

diff --git a/proposal.ipynb b/proposal.ipynb
deleted file mode 100644
index 927b95d..0000000
--- a/proposal.ipynb
+++ /dev/null
@@ -1,413 +0,0 @@
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-%% main.tex
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-\documentclass{article}
-\usepackage{amsfonts, amsmath, amsthm}
-\usepackage[letterpaper, total={6in, 9in}]{geometry}
-\usepackage[noend]{algorithmic}
-\usepackage[vlined,ruled,linesnumbered]{algorithm2e}
-\usepackage[pagebackref=true,breaklinks=true,letterpaper=true,colorlinks,bookmarks=false]{hyperref}
-
-\title{Multi-Agent Reinforcement Learning for Real-Time Frequency Regulation in Power Grids}
-\author{Your Name}
-\date{\today}
-
-\begin{document}
-\maketitle
-
-\input{sections/abstract.tex}
-\input{sections/introduction.tex}
-\input{sections/related-work.tex}
-\input{sections/formulation.tex}
-\input{sections/method.tex}
-\input{sections/experiments.tex}
-\input{sections/conclusion.tex}
-
-\clearpage
-\begin{center}
-    {\Large\bf Appendix}
-\end{center}
-\appendix
-\input{sections/app-proof-convergence.tex}
-
-\bibliographystyle{plain}
-\bibliography{refs}
-
-\end{document}
-
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-%% sections/abstract.tex
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-\begin{abstract}
-Renewable energy integration is disrupting power grid frequency control, creating faster dynamics and higher regulation costs. This project investigates multi-agent reinforcement learning for coordinating distributed controllable resources (batteries, generators, demand response) to maintain grid frequency at 60 Hz. We formulate the problem as a cooperative multi-agent MDP with continuous spaces, partial observability, and safety constraints. We compare three coordination approaches: centralized training with decentralized execution (MADDPG), value factorization (QMIX), and independent learning (IDDPG). All methods incorporate safety layers for constraint satisfaction. Evaluation on an IEEE 68-bus system aims to demonstrate 25\% cost reduction while maintaining frequency within $\pm$0.2 Hz for 99\% of operation.
-\end{abstract}
-
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-%% sections/introduction.tex
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-\section{Introduction}
-\label{sec:intro}
-
-\subsection{Motivation}
-Power grids must maintain frequency at 60 Hz by balancing generation and load. Renewable energy (solar/wind) now comprises 30\%+ of generation but provides no mechanical inertia, causing faster frequency dynamics and contributing to recent blackouts \cite{nerc2023}. Grid operators spend \$10B+ annually on frequency regulation. We need adaptive, data-driven control strategies.
-
-\subsection{Problem Statement}
-We consider a transmission grid with $N=20$ controllable units (batteries, gas generators, demand response) that must collectively stabilize frequency despite:
-\begin{itemize}
-    \item Stochastic renewable generation
-    \item Time-varying loads
-    \item Equipment outages (N-1 contingencies)
-    \item Partial observability and communication delays
-\end{itemize}
-
-This is a \textbf{sequential decision making problem} where agents must coordinate to balance frequency stability, cost, and safety constraints.
-
-\subsection{Why Multi-Agent RL?}
-Traditional PI controllers cannot optimize multi-step costs. Model predictive control requires accurate models that degrade over time. Centralized RL faces scalability issues (exponential action space) and single points of failure. Multi-agent RL enables distributed control but introduces coordination challenges: non-stationarity, credit assignment, and emergent cooperation.
-
-\subsection{Contributions}
-\begin{enumerate}
-    \item Detailed multi-agent MDP formulation with safety constraints
-    \item Comparison of coordination mechanisms (CTDE vs. independent learning)
-    \item Safety layers adapted to multi-agent settings
-    \item Evaluation on realistic power system with validated dynamics
-\end{enumerate}
-
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-%% sections/related-work.tex
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-\section{Related Work}
-\label{sec:related}
-
-\subsection{Power System Control}
-Traditional automatic generation control (AGC) uses PI controllers with fixed gains \cite{kundur1994}. Model predictive control improves performance but requires accurate system models that are expensive to maintain \cite{mohamed2012,venkat2008}.
-
-\subsection{RL for Power Systems}
-Deep RL has been applied to voltage control \cite{zhang2020} and economic dispatch \cite{cao2020}, but mostly assumes centralized control with full observability. Few works address frequency regulation specifically, and none systematically compare multi-agent coordination mechanisms with safety guarantees.
-
-\subsection{Safe RL}
-Safety-critical control requires constraint satisfaction. Approaches include constrained MDPs \cite{achiam2017}, safety layers that project actions onto safe sets \cite{dalal2018}, and shielding with formal guarantees \cite{alshiekh2018}.
-
-\subsection{Multi-Agent RL}
-Independent learners suffer from non-stationarity \cite{tan1993}. Centralized training with decentralized execution (CTDE) stabilizes learning: MADDPG uses centralized critics \cite{lowe2017}, while QMIX uses value factorization \cite{rashid2018}.
-
-\subsection{Gap We Address}
-No prior work systematically compares modern MARL algorithms on realistic frequency regulation with explicit safety constraints, validated dynamics, and evaluation on contingencies. Our work fills this gap.
-
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-%% sections/formulation.tex
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-\section{Problem Formulation}
-\label{sec:formulation}
-
-\subsection{Multi-Agent MDP}
-We define a cooperative MA-MDP:
-\begin{equation}
-\mathcal{M} = \langle \mathcal{N}, \mathcal{S}, \{\mathcal{A}^i\}_{i=1}^N, P, R, \gamma \rangle
-\end{equation}
-
-\textbf{Agents}: $\mathcal{N} = \{1, \ldots, 20\}$ controllable units (5 batteries, 8 gas plants, 7 demand response)
-
-\textbf{Global state} $s_t \in \mathcal{S} \subseteq \mathbb{R}^{140}$:
-\begin{itemize}
-    \item Frequency at 68 buses: $\mathbf{f}_t \in [59.5, 60.5]^{68}$ Hz
-    \item Generator outputs: $\mathbf{P}_t \in \mathbb{R}^{20}$ MW
-    \item Renewable generation: $P_t^{\text{solar}}, P_t^{\text{wind}}$
-    \item Load: $P_t^{\text{load}} \in [2000, 5000]$ MW
-    \item Time features (hour, day)
-\end{itemize}
-
-\textbf{Local observation} $o_t^i \in \mathcal{O}^i \subseteq \mathbb{R}^{15}$ (partial observability):
-\begin{itemize}
-    \item Local frequency $f_{\text{local}}^i$
-    \item Own output $P^i$ and capacity $(P_{\max}^i - P^i)$
-    \item System frequency deviation $\Delta f_{\text{sys}} = \frac{1}{68}\sum_k (f_k - 60)$ (broadcast)
-    \item Renewable forecast (4 steps ahead)
-\end{itemize}
-
-\textbf{Action} $a_t^i = \Delta P^i(t) \in [-\Delta P_{\max}^i, \Delta P_{\max}^i]$ MW/min
-
-\textbf{Constraints}:
-\begin{align}
-P_{\min}^i &\leq P^i + \Delta P^i \leq P_{\max}^i \quad \text{(capacity)} \\
-|\Delta P^i| &\leq R_{\max}^i \quad \text{(ramp rate)}
-\end{align}
-
-\textbf{Dynamics}: Governed by swing equation
-\begin{equation}
-2H \frac{df_k}{dt} = P_{\text{gen},k} - P_{\text{load},k} - \sum_{l} \frac{D_{kl}(f_k - f_l)}{X_l}
-\end{equation}
-plus stochastic disturbances (load, renewable forecast errors, N-1 outages). Dynamics are \textbf{unknown} to agents.
-
-\textbf{Shared reward}:
-\begin{equation}
-R(s, \mathbf{a}) = -\lambda_{\text{freq}} \sum_k (f_k - 60)^2 - \lambda_{\text{cost}} \sum_i C_i |\Delta P^i| - \lambda_{\text{wear}} \sum_i W_i(|\Delta P^i|)
-\end{equation}
-with $\lambda_{\text{freq}} = 1000$, $\lambda_{\text{cost}} = 1$, $\lambda_{\text{wear}} = 0.1$, and large penalty for violations $|f_k - 60| > 0.5$ Hz.
-
-\textbf{Objective}: Find policies $\{\pi^i\}_{i=1}^N$ that maximize
-\begin{equation}
-J = \mathbb{E}\left[\sum_{t=0}^\infty \gamma^t R(s_t, \mathbf{a}_t)\right]
-\end{equation}
-subject to safety: $\Pr[|f_k - 60| > 0.5] < 0.01$ for all $k$.
-
-Discount: $\gamma = 0.95$ (20-minute horizon).
-
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-%% sections/method.tex
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-\section{Proposed Methods}
-\label{sec:method}
-
-\subsection{Multi-Agent DDPG (MADDPG)}
-MADDPG \cite{lowe2017} uses centralized training with decentralized execution.
-
-\textbf{Decentralized actors}: Each agent has policy $\pi^i(o^i; \theta^i): \mathcal{O}^i \to \mathcal{A}^i$
-
-\textbf{Centralized critics}: $Q^i(s, a^1, \ldots, a^N; \phi^i)$ uses global state and all actions during training
-
-\textbf{Training}: 
-\begin{itemize}
-    \item Critic minimizes TD error: $\mathcal{L}(\phi^i) = \mathbb{E}[(Q^i(s, \mathbf{a}) - (R + \gamma Q^i(s', \mathbf{a}')))^2]$
-    \item Actor maximizes Q via gradient: $\nabla_{\theta^i} J = \mathbb{E}[\nabla_{\theta^i} \pi^i \nabla_{a^i} Q^i|_{a^i=\pi^i}]$
-\end{itemize}
-
-\textbf{Execution}: Only actors deployed using local observations (no communication needed)
-
-\subsection{QMIX}
-QMIX \cite{rashid2018} learns factorized Q-functions. We adapt to continuous actions via NAF:
-\begin{equation}
-Q^i(o^i, a^i) = V^i(o^i) - \frac{1}{2}(a^i - \mu^i(o^i))^2 P^i(o^i)
-\end{equation}
-
-Mixing network combines local Q-values monotonically:
-\begin{equation}
-Q_{\text{tot}}(s, \mathbf{a}) = g(Q^1, \ldots, Q^N, s)
-\end{equation}
-with $\partial Q_{\text{tot}}/\partial Q^i \geq 0$ ensuring individual-global-max property.
-
-\subsection{Independent DDPG (IDDPG)}
-Baseline where each agent learns independently. Expected to underperform due to non-stationarity.
-
-\subsection{Safety Mechanisms}
-
-\textbf{Action projection}: Project policy output onto constraint set
-\begin{equation}
-a^i = \arg\min_{a \in \mathcal{C}^i} \|\tilde{a}^i - a\|^2
-\end{equation}
-where $\mathcal{C}^i$ encodes capacity and ramp constraints.
-
-\textbf{Safety critic}: Learn $Q_{\text{safe}}^i(o^i, a^i)$ predicting violation probability. Bias policy toward safe regions during execution.
-
-\textbf{Lagrangian}: For soft constraints, learn dual variables:
-\begin{equation}
-\mathcal{L} = J(\theta) + \lambda(\mathbb{E}[\text{violation}] - \delta)
-\end{equation}
-
-All methods use these safety layers.
-
-\subsection{Algorithm}
-\begin{algorithm}[H]
-\caption{MADDPG with Safety}
-\For{episode $= 1, 2, \ldots$}{
-    Reset environment, get observations $\{o_0^i\}$\;
-    \For{$t = 0, 1, \ldots, T-1$}{
-        Each agent: $\tilde{a}_t^i = \pi^i(o_t^i) + \text{noise}$\;
-        Project: $a_t^i = \text{proj}_{\mathcal{C}^i}(\tilde{a}_t^i)$\;
-        Execute $\mathbf{a}_t$, observe $R_t, s_{t+1}, \{o_{t+1}^i\}$\;
-        Store transition in replay buffer\;
-    }
-    Sample minibatch, update critics and actors\;
-    Soft update target networks\;
-}
-\end{algorithm}
-
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-%% sections/experiments.tex
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-\section{Experimental Evaluation}
-\label{sec:experiments}
-
-\subsection{Environment}
-\textbf{System}: IEEE 68-bus transmission network \\
-\textbf{Simulator}: Pandapower + custom swing dynamics \\
-\textbf{Data}: ERCOT historical data (6 months), NREL renewable profiles \\
-\textbf{Episode}: 1440 steps (24 hours at 1-min resolution)
-
-\subsection{Baselines}
-\begin{itemize}
-    \item \textbf{B1: PI-AGC}: Industry standard proportional-integral controller
-    \item \textbf{B2: MPC}: Model predictive control (oracle with perfect model)
-    \item \textbf{B3: Behavioral cloning}: Supervised learning on historical data
-\end{itemize}
-
-\subsection{Proposed Methods}
-\begin{itemize}
-    \item \textbf{M1: MADDPG}: Centralized training, decentralized execution
-    \item \textbf{M2: QMIX}: Value factorization with NAF
-    \item \textbf{M3: IDDPG}: Independent learners (baseline)
-\end{itemize}
-
-\subsection{Training}
-\begin{itemize}
-    \item 5M steps (≈3,500 episodes) across 32 parallel environments
-    \item Learning rates: $\alpha_\pi = 10^{-4}$, $\alpha_Q = 10^{-3}$
-    \item Batch size: 256, Replay buffer: 500K
-    \item 10 random seeds, report mean $\pm$ 95\% CI
-\end{itemize}
-
-\subsection{Evaluation Scenarios}
-\begin{itemize}
-    \item \textbf{S1}: Normal operation (100 episodes)
-    \item \textbf{S2}: N-1 contingencies (50 episodes) - generator/line outages
-    \item \textbf{S3}: Renewable ramps (30 episodes) - extreme forecast errors
-    \item \textbf{S4}: Distribution shift (50 episodes) - different season
-\end{itemize}
-
-\subsection{Metrics}
-\textbf{Primary}:
-\begin{itemize}
-    \item Cumulative reward $\sum_t R_t$
-    \item Frequency stability: \% time with $|f - 60| < 0.2$ Hz (target: >99\%)
-    \item Regulation cost (total \$)
-    \item Constraint violations: \% time $|f - 60| > 0.5$ Hz (target: <1\%)
-\end{itemize}
-
-\textbf{Secondary}: Area control error, max deviation, renewable utilization
-
-\subsection{Ablations}
-\begin{itemize}
-    \item Coordination: MADDPG vs. QMIX vs. IDDPG
-    \item Observations: Full state vs. local+broadcast vs. local only
-    \item Safety: No safety vs. projection vs. critic vs. both
-    \item Reward weights: Vary $\lambda_{\text{freq}}, \lambda_{\text{cost}}$
-\end{itemize}
-
-\subsection{Success Criteria}
-\textbf{Tier 1 (minimum)}: MADDPG achieves >99\% frequency stability, >15\% cost reduction, zero critical failures
-
-\textbf{Tier 2 (expected)}: MADDPG/QMIX outperform IDDPG by >15\%, handle N-1 contingencies, >25\% cost reduction
-
-\textbf{Tier 3 (stretch)}: Achieve 90\% of MPC performance, generalize to larger grids
-
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-%% sections/conclusion.tex
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-\section{Conclusion}
-\label{sec:conclusion}
-
-This project provides the first systematic comparison of MARL algorithms for power grid frequency regulation with safety constraints. We expect to demonstrate that coordinated multi-agent RL (MADDPG/QMIX) can achieve significant cost savings (25\%+) over traditional control while maintaining reliability and handling contingencies.
-
-\textbf{Future work}: Scale to larger grids (100+ buses), sim-to-real transfer, adversarial robustness, integration with electricity markets.
-
-\textbf{Broader impacts}: Enabling higher renewable penetration reduces carbon emissions and costs. However, deployment requires careful consideration of cybersecurity, accountability, and equitable access.
-
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-%% sections/app-proof-convergence.tex
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-\section{Hyperparameters and Implementation}
-
-\subsection{Network Architectures}
-\textbf{Actor}: 3-layer MLP [15, 256, 128, 64, 1] with ReLU, tanh output \\
-\textbf{Critic}: 4-layer MLP [160, 512, 256, 128, 1] with ReLU
-
-\subsection{Key Hyperparameters}
-\begin{itemize}
-    \item Discount: $\gamma = 0.95$
-    \item Target update: $\tau = 0.01$
-    \item Exploration: OU noise $\sigma = 0.2 \to 0.05$ (linear decay)
-    \item Gradient clipping: norm = 0.5
-\end{itemize}
-
-\subsection{Computational Requirements}
-Training: ≈200 GPU-hours per method on 4 NVIDIA A100s \\
-Total: ≈1000 GPU-hours (feasible with institutional cluster)
-
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-%% refs.bib (create this file)
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-% Sample references - replace with actual citations
-
-@techreport{nerc2023,
-  author = {NERC},
-  title = {Frequency Response Initiative Report},
-  institution = {North American Electric Reliability Corporation},
-  year = {2023}
-}
-
-@book{kundur1994,
-  author = {Kundur, P.},
-  title = {Power System Stability and Control},
-  publisher = {McGraw-Hill},
-  year = {1994}
-}
-
-@article{mohamed2012,
-  author = {Mohamed, T. H. and others},
-  title = {Decentralized model predictive based load frequency control},
-  journal = {Energy Conversion and Management},
-  year = {2012}
-}
-
-@article{venkat2008,
-  author = {Venkat, A. N. and others},
-  title = {Distributed MPC strategies for automatic generation control},
-  journal = {IEEE Transactions on Control Systems Technology},
-  year = {2008}
-}
-
-@article{zhang2020,
-  author = {Zhang, Y. and others},
-  title = {Deep reinforcement learning based volt-VAR optimization},
-  journal = {IEEE Transactions on Smart Grid},
-  year = {2020}
-}
-
-@article{cao2020,
-  author = {Cao, D. and others},
-  title = {Reinforcement learning and its applications in power systems},
-  journal = {Journal of Modern Power Systems and Clean Energy},
-  year = {2020}
-}
-
-@inproceedings{achiam2017,
-  author = {Achiam, J. and others},
-  title = {Constrained Policy Optimization},
-  booktitle = {ICML},
-  year = {2017}
-}
-
-@article{dalal2018,
-  author = {Dalal, G. and others},
-  title = {Safe exploration in continuous action spaces},
-  journal = {arXiv:1801.08757},
-  year = {2018}
-}
-
-@inproceedings{alshiekh2018,
-  author = {Alshiekh, M. and others},
-  title = {Safe reinforcement learning via shielding},
-  booktitle = {AAAI},
-  year = {2018}
-}
-
-@inproceedings{tan1993,
-  author = {Tan, M.},
-  title = {Multi-agent reinforcement learning},
-  booktitle = {ICML},
-  year = {1993}
-}
-
-@inproceedings{lowe2017,
-  author = {Lowe, R. and others},
-  title = {Multi-agent actor-critic for mixed cooperative-competitive environments},
-  booktitle = {NeurIPS},
-  year = {2017}
-}
-
-@inproceedings{rashid2018,
-  author = {Rashid, T. and others},
-  title = {QMIX: Monotonic value function factorisation},
-  booktitle = {ICML},
-  year = {2018}
-}

From 80b4afe5841e6476530d78bc4c0bc3d09e6d2d7b Mon Sep 17 00:00:00 2001
From: Derek Smith <60520818+Br3ady@users.noreply.github.com>
Date: Sat, 1 Nov 2025 15:03:19 -0400
Subject: [PATCH 07/17] Delete sections directory

---
 sections/abstract.tex              |  5 -----
 sections/app-proof-convergence.tex |  7 ------
 sections/conclusion.tex            |  5 -----
 sections/experiments.tex           | 32 --------------------------
 sections/formulation.tex           | 11 ---------
 sections/introduction.tex          | 16 -------------
 sections/method.tex                | 36 ------------------------------
 sections/related-work.tex          |  6 -----
 8 files changed, 118 deletions(-)
 delete mode 100644 sections/abstract.tex
 delete mode 100644 sections/app-proof-convergence.tex
 delete mode 100644 sections/conclusion.tex
 delete mode 100644 sections/experiments.tex
 delete mode 100644 sections/formulation.tex
 delete mode 100644 sections/introduction.tex
 delete mode 100644 sections/method.tex
 delete mode 100644 sections/related-work.tex

diff --git a/sections/abstract.tex b/sections/abstract.tex
deleted file mode 100644
index e1d8e8b..0000000
--- a/sections/abstract.tex
+++ /dev/null
@@ -1,5 +0,0 @@
-%!TEX root = ../main.tex
-
-\begin{abstract}
-    Brief description of the problem, your solution, and the results.
-\end{abstract}
\ No newline at end of file
diff --git a/sections/app-proof-convergence.tex b/sections/app-proof-convergence.tex
deleted file mode 100644
index 1c6e269..0000000
--- a/sections/app-proof-convergence.tex
+++ /dev/null
@@ -1,7 +0,0 @@
-%!TEX root = ../main.tex
-
-\section{Proof of Theorem \ref{thm:convergence} }
-\label{sec:app-proof-convergence}
-\begin{proof}
-    Here is my proof.
-\end{proof}
\ No newline at end of file
diff --git a/sections/conclusion.tex b/sections/conclusion.tex
deleted file mode 100644
index 15bd700..0000000
--- a/sections/conclusion.tex
+++ /dev/null
@@ -1,5 +0,0 @@
-%!TEX root = ../main.tex
-\section{Conclusions}
-\label{sec:conclusions}
-
-Conclusions and potentially future research directions.
\ No newline at end of file
diff --git a/sections/experiments.tex b/sections/experiments.tex
deleted file mode 100644
index 265c11b..0000000
--- a/sections/experiments.tex
+++ /dev/null
@@ -1,32 +0,0 @@
-%!TEX root = ../main.tex
-\section{Experiments}
-\label{sec:experiments}
-
-How does your algorithm work and compare with others? What insights have you gained from the experiments? Use figures (see Figure \ref{fig:1} for example) and tables (see Table \ref{tab:1} for example) to better present your results.
-
-\begin{figure}[ht!]
-\centering
-\includegraphics[width=0.75\textwidth]{figure/SEASLogo.pdf}
-\label{fig:1}
-\caption{This is the caption for the figure.}
-\end{figure}
-
-\begin{table}[ht!]
-\centering
-\begin{tabular}{|c|c c c|} 
- \hline
-  &Col1 & Col2 & Col3 \\ [0.5ex] 
- \hline
- 1 & 6 & 87837 & 787 \\ 
- 2 & 7 & 78 & 5415 \\
- 3 & 545 & 778 & 7507 \\[1ex] 
- \hline
-\end{tabular}
-\caption{This is the caption for the table.}
-\label{tab:1}
-\end{table}
-
-\subsection{First case}
-
-Use subsections to separate different experiment settings, if needed.
-
diff --git a/sections/formulation.tex b/sections/formulation.tex
deleted file mode 100644
index 9519756..0000000
--- a/sections/formulation.tex
+++ /dev/null
@@ -1,11 +0,0 @@
-%!TEX root = ../main.tex
-\section{Problem Formulation}
-\label{sec:formulation}
-
-Mathematical formulation of your problem. For example
-\bea 
-\min_{u(t)} & \displaystyle \int_{0}^T g(x,u) dt + h(x(T)) \label{eq:1} \\
-\subject & \dot{x} = f(x,u)
-\eea 
-
-Equation \eqref{eq:1} can be referred by setting up its label.
\ No newline at end of file
diff --git a/sections/introduction.tex b/sections/introduction.tex
deleted file mode 100644
index ea097f1..0000000
--- a/sections/introduction.tex
+++ /dev/null
@@ -1,16 +0,0 @@
-%!TEX root = ../main.tex
-\section{Introduction}
-\label{sec:intro}
-
-Introduce your work. 
-\begin{itemize}
-\item Why is it important? 
-\item What's the motivation?
-\item What's your novelty?
-\end{itemize}
-
-Numbered list also helps summarizing the main contributions.
-\begin{enumerate}
-\item First point.
-\item Second point.
-\end{enumerate}
\ No newline at end of file
diff --git a/sections/method.tex b/sections/method.tex
deleted file mode 100644
index c24dc68..0000000
--- a/sections/method.tex
+++ /dev/null
@@ -1,36 +0,0 @@
-%!TEX root = ../main.tex
-\section{Solution Method}
-\label{sec:method}
-
-Describe your algorithm and its properties.
-
-\begin{algorithm}
-\label{alg:1}
-\caption{Calculate $y = x^n$}
-\begin{algorithmic} 
-\REQUIRE $n \geq 0 \vee x \neq 0$
-\ENSURE $y = x^n$
-\STATE $y \leftarrow 1$
-\IF{$n < 0$}
-\STATE $X \leftarrow 1 / x$
-\STATE $N \leftarrow -n$
-\ELSE
-\STATE $X \leftarrow x$
-\STATE $N \leftarrow n$
-\ENDIF
-\WHILE{$N \neq 0$}
-\IF{$N$ is even}
-\STATE $X \leftarrow X \times X$
-\STATE $N \leftarrow N / 2$
-\ELSE[$N$ is odd]
-\STATE $y \leftarrow y \times X$
-\STATE $N \leftarrow N - 1$
-\ENDIF
-\ENDWHILE
-\end{algorithmic}
-\end{algorithm}
-
-\begin{theorem}[Convergence]
-    \label{thm:convergence}
-    My Algorithm \ref{alg:1} is good.
-\end{theorem}
\ No newline at end of file
diff --git a/sections/related-work.tex b/sections/related-work.tex
deleted file mode 100644
index 1c0d0a7..0000000
--- a/sections/related-work.tex
+++ /dev/null
@@ -1,6 +0,0 @@
-%!TEX root = ../main.tex
-\section{Related Work}
-\label{sec:related-work}
-
-Cite some important works in this section, for example, ~\cite{bertsekas12book-optimalcontrol}. 
-Compare them with your project. One can also combine this section with Section \ref{sec:intro}.
\ No newline at end of file

From 33f72012309355deb423a35c1c736da347675066 Mon Sep 17 00:00:00 2001
From: Derek Smith <60520818+Br3ady@users.noreply.github.com>
Date: Sat, 1 Nov 2025 15:03:28 -0400
Subject: [PATCH 08/17] Delete draft-1.tex

---
 draft-1.tex | 464 ----------------------------------------------------
 1 file changed, 464 deletions(-)
 delete mode 100644 draft-1.tex

diff --git a/draft-1.tex b/draft-1.tex
deleted file mode 100644
index 5ce001f..0000000
--- a/draft-1.tex
+++ /dev/null
@@ -1,464 +0,0 @@
-# Project Proposal: Multi-Agent Reinforcement Learning for Real-Time Frequency Regulation in Power Grids with Renewable Integration
-
-## 1. Relevance to the Course (Sequential Decision Making in Dynamic Environments)
-
-This project addresses distributed optimal control in power system frequency regulation, formulated as a **Multi-Agent Markov Decision Process (MA-MDP)** where:
-
-- **Decision-makers**: Multiple controllable generation units (batteries, natural gas peakers, demand response aggregators) that must coordinate to maintain grid frequency at 60 Hz
-- **Dynamics**: Grid frequency evolves according to swing equations—determined by power balance between generation and load. Each agent's action affects system-wide frequency through coupled electromechanical dynamics:
-  ```
-  df/dt = (P_gen - P_load - P_losses) / (2H·S_base)
-  ```
-  where H is system inertia, and each agent contributes to total P_gen
-- **Sequential nature**: Control decisions must anticipate renewable generation forecasts (solar/wind ramps), load fluctuations, and other agents' actions. Incorrect responses cause cascading frequency deviations that take minutes to correct, requiring multi-step lookahead planning
-
-The problem exhibits core RL characteristics: continuous high-dimensional state/action spaces, partial observability (agents see local measurements with communication delays), stochastic disturbances (renewable intermittency, unexpected load changes), and critical safety constraints (frequency must stay within ±0.5 Hz to prevent equipment damage and blackouts). The **multi-agent coordination** aspect introduces challenges beyond single-agent RL: credit assignment across agents, non-stationarity from simultaneously learning policies, and scalability to realistic grid sizes (50+ agents).
-
----
-
-## 2. Motivation and Related Work
-
-### Why This Matters
-
-The rapid integration of renewable energy fundamentally disrupts power grid operations. Solar and wind now comprise >30% of generation in many regions [EIA, 2024], but their intermittency creates severe frequency control challenges. Traditional synchronous generators provide natural inertia that stabilizes frequency; renewables (inverter-based) do not. This "low-inertia" problem causes:
-- **Faster frequency dynamics**: Rate of change of frequency (RoCoF) has doubled in some grids [NERC, 2023]
-- **Increased regulation costs**: Grid operators pay $10B+ annually for frequency services in the U.S. alone
-- **Reliability risks**: Several major blackouts (South Australia 2016, Texas 2021) linked to inadequate frequency response
-
-Current solutions rely on costly fast-response reserves or curtail renewable generation. **Data-driven multi-agent RL** offers coordinated, adaptive control that learns optimal dispatch strategies from operational data, potentially reducing costs by 20-40% while improving reliability [Venkat et al., 2022].
-
-### Challenge
-
-The core technical barriers:
-
-1. **Safety-critical real-time control**: Frequency deviations >1 Hz trigger automatic load shedding (blackouts). Controllers must guarantee constraint satisfaction, not just optimize in expectation.
-
-2. **Scalability**: Realistic transmission systems have 100+ controllable units. Centralized control suffers from communication bottlenecks and single points of failure; fully decentralized control leads to suboptimal coordination.
-
-3. **Partial observability and delays**: Agents observe local bus frequency/voltage with 2-4 second telemetry delays. True system state (all bus frequencies, line flows) is not directly observable.
-
-4. **Non-stationary environment**: Load patterns shift hourly/seasonally, renewable output depends on weather, and topology changes due to line/generator outages. Policies must adapt online without catastrophic forgetting.
-
-5. **Multi-objective tradeoffs**: Minimize regulation cost AND frequency deviation AND wear on equipment (battery cycling, generator ramping) while respecting power limits and reserve requirements.
-
-### Prior Work
-
-**1. Classical control approaches**:
-- Automatic Generation Control (AGC) with PI controllers [Kundur, 1994]: Industry standard but cannot handle nonlinear dynamics or optimize multi-step costs
-- Model Predictive Control (MPC) [Mohamed et al., 2012; Venkat et al., 2008]: Effective but requires accurate system models (costly to obtain, degrade with topology changes)
-
-**2. Single-agent RL for power systems**:
-- DQN/DDPG for generator dispatch [Zhang et al., 2020; Cao et al., 2020]: Proof-of-concept on small test systems (IEEE 14-bus), but centralized architecture doesn't scale
-- Safe RL with constraints [Dalal et al., 2018 applied to power]: Barrier functions and Lagrangian methods, but limited to single controller
-
-**3. Multi-agent RL foundations**:
-- Independent learners (IQL) [Tan, 1993]: Each agent learns separately; suffers from non-stationary environment
-- Centralized training with decentralized execution (CTDE) [Lowe et al., 2017; Foerster et al., 2018]: MADDPG, QMIX enable coordination via centralized critic during training
-- Communication protocols [Sukhbaatar et al., 2016; Jiang & Lu, 2018]: CommNet, TarMAC allow explicit message passing between agents
-
-**4. Power systems + MARL (limited prior work)**:
-- [Zhou et al., 2020]: MADDPG for voltage control in distribution networks—different problem (voltage vs. frequency)
-- [Huang et al., 2021]: Multi-agent dispatch for economic optimization, but ignores frequency dynamics and safety constraints
-- [Liu et al., 2023]: Independent Q-learning for AGC, but no comparison to coordinated methods or constraint handling
-
-### Gap We Address
-
-**No prior work systematically evaluates modern MARL algorithms (CTDE vs. communication-based vs. independent) on realistic frequency regulation with explicit safety constraints and renewable integration.** Specifically:
-
-1. Compare coordination mechanisms: CTDE (MADDPG, QMIX) vs. learned communication (TarMAC) vs. independent learners
-2. Develop constraint-aware training using safety layers [Dalal et al., 2018] adapted to multi-agent setting
-3. Evaluate generalization to topology changes (N-1 contingencies) and distribution shift (weather patterns)
-4. Use realistic power system simulator (PYPOWER/Pandapower) with validated models, not toy grids
-
----
-
-## 3. Problem Definition
-
-### Agent and Environment
-
-- **Agents**: N = 20 controllable units in a transmission grid:
-  - 5 battery energy storage systems (BESS)
-  - 8 natural gas peaker plants (fast-response generators)
-  - 7 demand response aggregators (controllable loads)
-- **Environment**: IEEE 68-bus transmission system with stochastic renewable generation and load
-
-### Formal Multi-Agent MDP Model
-
-We define a cooperative MA-MDP: **M = (S, {A^i}, P, {R^i}, γ, N)** with shared reward (team objective).
-
-**Global state space S ⊆ ℝ^m** (continuous, partially observable):
-- Grid frequency at each bus: f_k ∈ [59.5, 60.5] Hz for k = 1...68
-- Active power output of all generators: P_g^j ∈ [P_min^j, P_max^j] MW
-- Renewable generation: P_solar ∈ [0, 500] MW, P_wind ∈ [0, 800] MW
-- Aggregate load: P_load ∈ [2000, 5000] MW
-- Time of day: t_hour ∈ [0, 24), t_day ∈ [0, 7) (sin/cos encoded)
-- Line flows: F_l ∈ [-F_max^l, F_max^l] MW for critical transmission lines
-
-**Local observation for agent i, O^i(s) ⊆ ℝ^{d_i}**:
-- Local bus frequency f_local
-- Own power output P^i, available capacity margin (P_max^i - P^i)
-- Renewable forecast (next 4 steps, 1-minute resolution)
-- Aggregate system frequency deviation (broadcast signal): Δf_sys = 1/68 ∑_k (f_k - 60)
-- Messages from neighboring agents (if communication enabled): m_j for j ∈ N(i)
-
-**Action space for agent i, A^i ⊆ ℝ**:
-- Change in power output: Δ P^i ∈ [-ΔP_max^i, ΔP_max^i] MW/min
-- Constraints:
-  - **Capacity limits**: P^i + Δ P^i ∈ [P_min^i, P_max^i]
-  - **Ramp rate limits**: |Δ P^i| ≤ R_max^i (batteries: 50 MW/min, gas: 10 MW/min, DR: 5 MW/min)
-  - **Battery energy limits**: State of charge ∈ [20%, 90%]
-
-**Joint action**: a = (a^1, ..., a^N) ∈ A = A^1 × ... × A^N
-
-**Transition dynamics P(s' | s, a)**: Governed by power flow equations + swing dynamics:
-
-1. **Swing equation** (frequency dynamics):
-   ```
-   2H · df_k/dt = P_gen,k - P_load,k - ∑_l D_kl(f_k - f_l) / X_l
-   ```
-   where D_kl is damping, X_l is line reactance
-
-2. **Power flow**: DC approximation for line flows (validated linearization for small frequency deviations)
-
-3. **Stochastic disturbances**:
-   - Load: Ornstein-Uhlenbeck process + hourly trend
-   - Solar: Forecast + prediction error (σ = 15% of forecast)
-   - Wind: Autoregressive model with weather-dependent variance
-
-4. **Contingency events**: Random generator/line outages (N-1) with probability 0.001 per timestep
-
-Dynamics are **partially unknown**: agents don't have access to full grid topology, line impedances, or precise load models. They must learn from interactions.
-
-**Shared reward function R(s, a)** (team objective):
-```
-R(s, a) = -λ_freq · ∑_k (f_k - 60)² 
-          - λ_cost · ∑_i C_i(P^i) · |Δ P^i|
-          - λ_wear · ∑_i W_i(|Δ P^i|)
-          - λ_violation · 𝟙[constraints violated]
-```
-
-**Terms**:
-- **Frequency deviation penalty**: λ_freq = 1000 (quadratic to heavily penalize large deviations)
-- **Regulation cost**: C_i(P) = marginal cost of power adjustment (batteries: $5/MWh, gas: $50/MWh, DR: $30/MWh)
-- **Wear cost**: W_i = degradation from cycling (batteries: $0.01/MW for state-of-health, gas: thermal stress)
-- **Constraint violation**: Large penalty (10,000) if frequency exceeds ±0.5 Hz or power limits violated
-
-**Discount factor γ = 0.95** (20 steps ≈ 20 minutes planning horizon at 1-min control intervals)
-
-### Assumptions
-
-1. **Cooperative agents**: All agents share the team objective (no adversarial behavior or misaligned incentives)
-2. **Partial observability**: Agents observe local state + limited broadcast info; cannot see full network state
-3. **Communication topology**: Agents can exchange messages with neighbors in electrical topology (graph structure), with 2-second delay
-4. **Unknown dynamics**: Grid model (admittance matrix, inertia constants) not provided; must be learned
-5. **Stochastic environment**: Renewable generation and load are random but satisfy historical distributions
-6. **Safety-critical constraints**: Hard constraints on frequency bounds; soft constraints on power limits (penalized but recoverable)
-
-### Available Data
-
-**Offline historical data** (for pre-training/validation):
-- 6 months of 1-minute resolution SCADA data from ERCOT (Texas grid)
-- Features: bus frequencies, generator outputs, load, renewable generation, prices
-- Behavior policy: Historical AGC PI controllers + manual dispatch
-- ~260,000 timesteps (limited compared to state-action space)
-
-**Simulation environment**:
-- **Power flow simulator**: Pandapower (Python) with IEEE 68-bus test system
-- **Dynamics simulator**: Custom swing equation integrator (Runge-Kutta 4th order)
-- **Renewable/load models**: Trained on historical data (NREL solar/wind profiles, ERCOT load)
-- **Contingency scenarios**: Library of 50 N-1 outage events (generator and line failures)
-
-**Software infrastructure**:
-- **MARL framework**: PyMARL2, EPyMARL (supports CTDE algorithms)
-- **RL libraries**: Stable-Baselines3 (for single-agent baselines), RLlib (distributed training)
-- **Deep learning**: PyTorch for custom network architectures
-- **Power system tools**: Pandapower, PYPOWER for steady-state validation
-- **Compute**: Multi-GPU training (4x NVIDIA A100) for parallel environment rollouts
-
----
-
-## 4. Proposed Method and Goals
-
-### Candidate Methods
-
-We will implement and compare four multi-agent approaches:
-
-#### 4.1 Multi-Agent Deep Deterministic Policy Gradient (MADDPG) [Primary CTDE Method]
-
-**Rationale**: State-of-the-art CTDE algorithm that addresses non-stationarity through centralized critic during training, decentralized actor during execution.
-
-**Architecture**:
-- **Decentralized actors**: π^i(a^i | o^i; θ^i) for each agent i (only uses local observation)
-- **Centralized critic**: Q(s, a^1, ..., a^N; φ) evaluates joint action given global state (available during training)
-- **Training**: Each actor optimizes using policy gradient with centralized Q:
-  ```
-  ∇_θ^i J = E[∇_θ^i π^i(o^i) · ∇_a^i Q(s, a^1, ..., a^N)|_{a^i = π^i(o^i)}]
-  ```
-- **Execution**: Only actors deployed (decentralized, no communication needed)
-
-**Network design**:
-- Actor: 3-layer MLP (256-128-64), tanh activation, outputs continuous action
-- Critic: 4-layer MLP (512-256-128-64), takes concatenated state and all actions
-- Target networks with soft updates (τ = 0.01)
-
-**Constraint handling**: Project actions onto feasible set (ramp rate, capacity limits) using quadratic programming layer
-
-#### 4.2 QMIX with Continuous Actions [Alternative CTDE Method]
-
-**Rationale**: Value-based CTDE that learns factorized Q-function Q_tot = g(Q^1, ..., Q^N) where g is a monotonic mixing network. Often more sample-efficient than policy gradients.
-
-**Architecture**:
-- **Local Q-networks**: Q^i(o^i, a^i; θ^i) for discrete action spaces
-- **Modification for continuous actions**: Use NAF (Normalized Advantage Functions) or discretize action space into 21 bins
-- **Mixing network**: Hypernetwork generates weights based on global state to combine local Q-values
-
-**Advantage**: Enables centralized training with decentralized execution without explicit policy, potentially more stable
-
-#### 4.3 Targeted Multi-Agent Communication (TarMAC) [Communication-Based Method]
-
-**Rationale**: Learn when and what to communicate. Agents exchange learned feature representations to improve coordination without requiring full state access.
-
-**Architecture**:
-- **Communication protocol**: 
-  ```
-  m^i_t = signature_net^i(o^i_t, h^i_{t-1})  // what to send
-  α^i_j = attention(h^i, m^j)                 // who to listen to
-  h^i_t = GRU(o^i_t, ∑_j α^i_j m^j)          // message aggregation
-  a^i_t = actor(h^i_t)                        // action selection
-  ```
-- **Signature network**: Learns compact message representation (dim=32)
-- **Attention mechanism**: Decides communication importance (scaled dot-product)
-
-**Training**: End-to-end with policy gradient; communication naturally learned to maximize return
-
-**Advantage**: More realistic for deployment (only requires neighbor communication, not full state access)
-
-#### 4.4 Independent Deep Deterministic Policy Gradient (IDDPG) [Baseline]
-
-**Rationale**: Each agent learns independently treating others as part of environment. Provides lower bound on coordination benefit.
-
-**Architecture**: Standard DDPG per agent with only local observation, no shared critic or communication
-
-**Expected weakness**: Non-stationarity from other agents updating policies, likely suboptimal coordination
-
-### Justification of Method Choices
-
-- **MADDPG**: Proven effective for continuous control with multiple agents; centralized critic stabilizes training
-- **QMIX**: Tests whether value factorization outperforms direct policy optimization in this domain
-- **TarMAC**: Evaluates learned communication vs. fixed CTDE; more practical for realistic deployment with communication constraints
-- **IDDPG**: Essential baseline to quantify coordination gains
-
-All methods will incorporate **safety layers** [Dalal et al., 2018]:
-- **Projection layer**: Projects policy output onto constraint set (capacity, ramp rates) via analytical solution
-- **Safety critic**: Separate Q-network trained to predict constraint violation probability
-- **Lagrangian relaxation**: For soft constraints, learn dual variables λ to penalize violations:
-  ```
-  L_safe = L_RL + ∑_c λ_c · max(0, constraint_violation_c)
-  ```
-
-### Concrete Goals and Success Criteria
-
-**Primary Objectives**:
-
-1. **Frequency stability**: Maintain |f_k - 60 Hz| < 0.2 Hz for 99% of time (vs. 95% with baseline AGC)
-2. **Cost reduction**: Achieve ≥25% reduction in regulation costs compared to PI-based AGC
-3. **Coordination efficiency**: MADDPG/QMIX outperform IDDPG by ≥15% in cumulative reward
-4. **Constraint satisfaction**: Zero critical violations (f outside [59.5, 60.5] Hz); <2% soft constraint violations (power limits)
-
-**Performance Metrics**:
-
-1. **Frequency regulation quality**:
-   - Area Control Error (ACE): Integral of frequency deviation over episode
-   - RoCoF: Maximum rate of change of frequency during disturbances
-   - Nadir frequency: Minimum frequency after contingency events
-
-2. **Economic metrics**:
-   - Total regulation cost: ∑_t ∑_i C_i(P^i_t) · |Δ P^i_t|
-   - Dispatch efficiency: Ratio of renewable energy utilized vs. curtailed
-
-3. **Safety metrics**:
-   - Constraint violation rate: % timesteps with violations
-   - Severity of worst violation: max_t |f_t - 60|
-
-4. **Learning metrics**:
-   - Sample efficiency: Reward vs. environment steps
-   - Convergence stability: Variance of returns in final 100 episodes
-
-5. **Coordination metrics** (for MARL):
-   - Communication overhead: Bits transmitted per agent per timestep (TarMAC)
-   - Credit assignment: Individual agent contribution via Shapley values
-
-**Success Criteria**:
-
-✅ **Tier 1 (Minimum viable)**: MADDPG achieves frequency stability goal (99% within ±0.2 Hz) and 15% cost reduction
-
-✅ **Tier 2 (Expected)**: MADDPG/QMIX outperform IDDPG by ≥15% and handle N-1 contingencies without blackouts
-
-✅ **Tier 3 (Stretch)**: TarMAC matches MADDPG performance with <20% communication bandwidth; policy generalizes to unseen contingency types
-
-### Evaluation Plan
-
-**Baselines**:
-
-1. **PI-based AGC** (industry standard): Proportional-integral controller with droop control
-2. **Centralized MPC** (upper bound): Model predictive control with perfect model (oracle)
-3. **IDDPG** (independent learners): Quantifies benefit of coordination
-4. **Behavioral cloning**: Supervised learning on historical ERCOT data (offline benchmark)
-
-**Training Protocol**:
-
-1. **Pre-training phase** (optional):
-   - Warm-start MADDPG with behavioral cloning on offline data (6 months)
-   - Initialize dynamics model for MPC baseline via system identification
-
-2. **Online training**:
-   - Parallel environment rollouts (32 environments, 4 GPUs)
-   - Episode length: 1440 steps (24 hours of operation)
-   - Total training: 5M timesteps (≈3,500 episodes)
-   - Replay buffer: 500k transitions
-   - Exploration: Ornstein-Uhlenbeck noise on actions (σ decreases linearly)
-
-3. **Curriculum learning**:
-   - Stage 1 (0-1M steps): Normal operation, no contingencies
-   - Stage 2 (1-3M steps): Introduce N-1 outages (5% of episodes)
-   - Stage 3 (3-5M steps): Extreme scenarios (high renewable variability, multiple outages)
-
-**Evaluation Scenarios**:
-
-1. **Normal operation**: 100 episodes with historical weather/load patterns (different months than training)
-2. **Contingency stress test**: 50 episodes with N-1 generator/line outages
-3. **Renewable ramp events**: 30 episodes with extreme solar/wind forecast errors (3σ events)
-4. **Topology changes**: 20 episodes with 10% fewer controllable generators (test graceful degradation)
-5. **Distribution shift**: Evaluate on different season (summer vs. winter), different grid size (scale to IEEE 118-bus)
-
-**Metrics Collection**:
-
-- **Per-episode**: Total reward, ACE, constraint violations, cost breakdown
-- **Per-timestep**: Frequency at all buses, line flows, agent actions
-- **Statistical testing**: 10 random seeds, report mean ± 95% confidence interval
-- **Visualization**: Frequency heatmaps, agent action timeseries, communication patterns (TarMAC)
-
-**Ablation Studies**:
-
-1. **Coordination mechanism**: MADDPG vs. QMIX vs. TarMAC vs. IDDPG
-2. **Observation space**: Full state (cheating) vs. local + broadcast vs. local only
-3. **Communication topology** (TarMAC): Electrical neighbors vs. k-hop neighbors vs. full connectivity
-4. **Safety mechanisms**: With vs. without safety critic and projection layers
-5. **Reward shaping**: Vary λ_freq, λ_cost, λ_wear to understand tradeoffs
-
-**Computational Requirements**:
-
-- **Training**: ~200 GPU-hours per algorithm (4 GPUs × 50 hours)
-- **Evaluation**: ~10 CPU-hours per method
-- **Total**: ~1000 GPU-hours for all methods + ablations (feasible with institutional cluster)
-
-### Feasibility and Risk Mitigation
-
-✅ **Highly Feasible**:
-- Pandapower is mature, well-documented power system simulator
-- PyMARL2/EPyMARL provide tested MARL implementations
-- IEEE test cases widely available and validated
-- Team has access to ERCOT historical data and compute cluster
-
-⚠️ **Risk 1: Training instability in multi-agent setting**
-- **Mitigation**: 
-  - Use gradient clipping (norm=0.5)
-  - Target network soft updates to reduce non-stationarity
-  - If instability persists, fall back to sequential training (update one agent at a time)
-
-⚠️ **Risk 2: Sim-to-real gap (if validating on real grid data)**
-- **Mitigation**: 
-  - Validate simulator against historical ERCOT data (frequency response to known events)
-  - Add noise to simulator (±5% parameter uncertainty) for robustness
-  - Domain randomization: Train on distribution of grid topologies
-
-⚠️ **Risk 3: Insufficient exploration in safety-critical setting**
-- **Mitigation**: 
-  - Use importance sampling to oversample rare critical states in replay buffer
-  - Safe exploration via optimistic constraint estimation [Berkenkamp et al., 2017]
-  - If exploration too dangerous, leverage offline pre-training more heavily
-
-⚠️ **Risk 4: Communication protocol learning may not converge (TarMAC)**
-- **Mitigation**: 
-  - Pre-train signature network with autoencoder objective (reconstruct observation from message)
-  - Regularize communication (entropy bonus to encourage information sharing)
-  - If fails, compare to hand-designed communication (broadcast frequency error)
-
-⚠️ **Risk 5: Scalability to larger grids (118+ buses)**
-- **Mitigation**: 
-  - Use graph neural networks (GNN) for agent policies to exploit grid topology structure
-  - Hierarchical control: Train area-level coordinators + local agents
-  - Focus on 68-bus for main results; 118-bus as stretch goal
-
-### Timeline (12 weeks)
-
-**Weeks 1-2**: Environment setup and baselines
-- Implement Pandapower simulation wrapper with swing dynamics
-- Code PI-AGC and MPC baselines
-- Validate simulator against ERCOT historical frequency response
-
-**Weeks 3-4**: Single-agent methods
-- Implement IDDPG baseline (independent learners)
-- Debug training loop, replay buffer, safety projections
-- Initial results on small 14-bus test system
-
-**Weeks 5-7**: Multi-agent methods
-- Implement MADDPG and QMIX
-- Scale to 68-bus system
-- Hyperparameter tuning (learning rates, replay buffer size, target update frequency)
-
-**Week 8**: Communication-based method
-- Implement TarMAC with attention mechanism
-- Experiment with communication topologies
-
-**Weeks 9-10**: Evaluation and ablations
-- Run full evaluation suite (100+ episodes per method)
-- Ablation studies (observation space, safety mechanisms, reward weights)
-- Statistical significance testing
-
-**Week 11**: Analysis and visualization
-- Generate plots (frequency trajectories, cost breakdown, learning curves)
-- Analyze failure modes (what causes constraint violations?)
-- Contingency analysis (N-1 response quality)
-
-**Week 12**: Report writing and presentation
-- Draft final report with all results
-- Create presentation slides
-- Prepare demo video (visualizing learned policies)
-
----
-
-## References
-
-**Power systems fundamentals**:
-- Kundur, P. (1994). *Power System Stability and Control*. McGraw-Hill.
-- NERC (2023). "Frequency Response Initiative Report." North American Electric Reliability Corporation.
-
-**Control and optimization**:
-- Mohamed, T. H., et al. (2012). "Decentralized model predictive based load frequency control in an interconnected power system." *Energy Conversion and Management*.
-- Venkat, A. N., et al. (2008). "Distributed MPC strategies for automatic generation control." *IEEE Transactions on Control Systems Technology*.
-
-**Multi-agent reinforcement learning**:
-- Tan, M. (1993). "Multi-agent reinforcement learning: Independent vs. cooperative agents." *ICML*.
-- Lowe, R., et al. (2017). "Multi-agent actor-critic for mixed cooperative-competitive environments." *NeurIPS*.
-- Foerster, J., et al. (2018). "Counterfactual multi-agent policy gradients." *AAAI*.
-- Rashid, T., et al. (2018). "QMIX: Monotonic value function factorisation for decentralized multi-agent reinforcement learning." *ICML*.
-- Sukhbaatar, S., et al. (2016). "Learning multiagent communication with backpropagation." *NeurIPS*.
-- Jiang, J., & Lu, Z. (2018). "Learning attentional communication for multi-agent cooperation." *NeurIPS*.
-
-**Safe reinforcement learning**:
-- Dalal, G., et al. (2018). "Safe exploration in continuous action spaces." *arXiv:1801.08757*.
-- Berkenkamp, F., et al. (2017). "Safe model-based reinforcement learning with stability guarantees." *NeurIPS*.
-
-**Power systems + RL/MARL**:
-- Zhang, Y., et al. (2020). "Deep reinforcement learning based volt-VAR optimization in smart distribution systems." *IEEE Transactions on Smart Grid*.
-- Cao, D., et al. (2020). "Reinforcement learning and its applications in modern power and energy systems: A review." *Journal of Modern Power Systems and Clean Energy*.
-- Zhou, S., et al. (2020). "Multi-agent reinforcement learning for coordinated voltage control." *IEEE PES General Meeting*.
-- Huang, Q., et al. (2021). "Multi-agent deep reinforcement learning for HVAC control in commercial buildings." *IEEE Transactions on Smart Grid*.
-- Liu, Y., et al. (2023). "Distributed reinforcement learning for decentralized automatic generation control." *Electric Power Systems Research*.
-- Venkat, D., et al. (2022). "Economic and reliability impacts of RL-based frequency regulation." *IEEE Transactions on Power Systems*.
-
-**Data sources**:
-- U.S. Energy Information Administration (EIA). (2024). "Electric Power Monthly."
-- ERCOT (Electric Reliability Council of Texas). SCADA historical data.
-- NREL (National Renewable Energy Laboratory). Solar and wind generation profiles.

From 5fbd4756c77913164752c6d41277159f66a9db44 Mon Sep 17 00:00:00 2001
From: Derek Smith <60520818+Br3ady@users.noreply.github.com>
Date: Sat, 1 Nov 2025 15:03:39 -0400
Subject: [PATCH 09/17] Delete main.tex

---
 main.tex | 413 -------------------------------------------------------
 1 file changed, 413 deletions(-)
 delete mode 100644 main.tex

diff --git a/main.tex b/main.tex
deleted file mode 100644
index 927b95d..0000000
--- a/main.tex
+++ /dev/null
@@ -1,413 +0,0 @@
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-%% main.tex
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-\documentclass{article}
-\usepackage{amsfonts, amsmath, amsthm}
-\usepackage[letterpaper, total={6in, 9in}]{geometry}
-\usepackage[noend]{algorithmic}
-\usepackage[vlined,ruled,linesnumbered]{algorithm2e}
-\usepackage[pagebackref=true,breaklinks=true,letterpaper=true,colorlinks,bookmarks=false]{hyperref}
-
-\title{Multi-Agent Reinforcement Learning for Real-Time Frequency Regulation in Power Grids}
-\author{Your Name}
-\date{\today}
-
-\begin{document}
-\maketitle
-
-\input{sections/abstract.tex}
-\input{sections/introduction.tex}
-\input{sections/related-work.tex}
-\input{sections/formulation.tex}
-\input{sections/method.tex}
-\input{sections/experiments.tex}
-\input{sections/conclusion.tex}
-
-\clearpage
-\begin{center}
-    {\Large\bf Appendix}
-\end{center}
-\appendix
-\input{sections/app-proof-convergence.tex}
-
-\bibliographystyle{plain}
-\bibliography{refs}
-
-\end{document}
-
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-%% sections/abstract.tex
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-\begin{abstract}
-Renewable energy integration is disrupting power grid frequency control, creating faster dynamics and higher regulation costs. This project investigates multi-agent reinforcement learning for coordinating distributed controllable resources (batteries, generators, demand response) to maintain grid frequency at 60 Hz. We formulate the problem as a cooperative multi-agent MDP with continuous spaces, partial observability, and safety constraints. We compare three coordination approaches: centralized training with decentralized execution (MADDPG), value factorization (QMIX), and independent learning (IDDPG). All methods incorporate safety layers for constraint satisfaction. Evaluation on an IEEE 68-bus system aims to demonstrate 25\% cost reduction while maintaining frequency within $\pm$0.2 Hz for 99\% of operation.
-\end{abstract}
-
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-%% sections/introduction.tex
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-\section{Introduction}
-\label{sec:intro}
-
-\subsection{Motivation}
-Power grids must maintain frequency at 60 Hz by balancing generation and load. Renewable energy (solar/wind) now comprises 30\%+ of generation but provides no mechanical inertia, causing faster frequency dynamics and contributing to recent blackouts \cite{nerc2023}. Grid operators spend \$10B+ annually on frequency regulation. We need adaptive, data-driven control strategies.
-
-\subsection{Problem Statement}
-We consider a transmission grid with $N=20$ controllable units (batteries, gas generators, demand response) that must collectively stabilize frequency despite:
-\begin{itemize}
-    \item Stochastic renewable generation
-    \item Time-varying loads
-    \item Equipment outages (N-1 contingencies)
-    \item Partial observability and communication delays
-\end{itemize}
-
-This is a \textbf{sequential decision making problem} where agents must coordinate to balance frequency stability, cost, and safety constraints.
-
-\subsection{Why Multi-Agent RL?}
-Traditional PI controllers cannot optimize multi-step costs. Model predictive control requires accurate models that degrade over time. Centralized RL faces scalability issues (exponential action space) and single points of failure. Multi-agent RL enables distributed control but introduces coordination challenges: non-stationarity, credit assignment, and emergent cooperation.
-
-\subsection{Contributions}
-\begin{enumerate}
-    \item Detailed multi-agent MDP formulation with safety constraints
-    \item Comparison of coordination mechanisms (CTDE vs. independent learning)
-    \item Safety layers adapted to multi-agent settings
-    \item Evaluation on realistic power system with validated dynamics
-\end{enumerate}
-
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-%% sections/related-work.tex
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-\section{Related Work}
-\label{sec:related}
-
-\subsection{Power System Control}
-Traditional automatic generation control (AGC) uses PI controllers with fixed gains \cite{kundur1994}. Model predictive control improves performance but requires accurate system models that are expensive to maintain \cite{mohamed2012,venkat2008}.
-
-\subsection{RL for Power Systems}
-Deep RL has been applied to voltage control \cite{zhang2020} and economic dispatch \cite{cao2020}, but mostly assumes centralized control with full observability. Few works address frequency regulation specifically, and none systematically compare multi-agent coordination mechanisms with safety guarantees.
-
-\subsection{Safe RL}
-Safety-critical control requires constraint satisfaction. Approaches include constrained MDPs \cite{achiam2017}, safety layers that project actions onto safe sets \cite{dalal2018}, and shielding with formal guarantees \cite{alshiekh2018}.
-
-\subsection{Multi-Agent RL}
-Independent learners suffer from non-stationarity \cite{tan1993}. Centralized training with decentralized execution (CTDE) stabilizes learning: MADDPG uses centralized critics \cite{lowe2017}, while QMIX uses value factorization \cite{rashid2018}.
-
-\subsection{Gap We Address}
-No prior work systematically compares modern MARL algorithms on realistic frequency regulation with explicit safety constraints, validated dynamics, and evaluation on contingencies. Our work fills this gap.
-
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-%% sections/formulation.tex
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-\section{Problem Formulation}
-\label{sec:formulation}
-
-\subsection{Multi-Agent MDP}
-We define a cooperative MA-MDP:
-\begin{equation}
-\mathcal{M} = \langle \mathcal{N}, \mathcal{S}, \{\mathcal{A}^i\}_{i=1}^N, P, R, \gamma \rangle
-\end{equation}
-
-\textbf{Agents}: $\mathcal{N} = \{1, \ldots, 20\}$ controllable units (5 batteries, 8 gas plants, 7 demand response)
-
-\textbf{Global state} $s_t \in \mathcal{S} \subseteq \mathbb{R}^{140}$:
-\begin{itemize}
-    \item Frequency at 68 buses: $\mathbf{f}_t \in [59.5, 60.5]^{68}$ Hz
-    \item Generator outputs: $\mathbf{P}_t \in \mathbb{R}^{20}$ MW
-    \item Renewable generation: $P_t^{\text{solar}}, P_t^{\text{wind}}$
-    \item Load: $P_t^{\text{load}} \in [2000, 5000]$ MW
-    \item Time features (hour, day)
-\end{itemize}
-
-\textbf{Local observation} $o_t^i \in \mathcal{O}^i \subseteq \mathbb{R}^{15}$ (partial observability):
-\begin{itemize}
-    \item Local frequency $f_{\text{local}}^i$
-    \item Own output $P^i$ and capacity $(P_{\max}^i - P^i)$
-    \item System frequency deviation $\Delta f_{\text{sys}} = \frac{1}{68}\sum_k (f_k - 60)$ (broadcast)
-    \item Renewable forecast (4 steps ahead)
-\end{itemize}
-
-\textbf{Action} $a_t^i = \Delta P^i(t) \in [-\Delta P_{\max}^i, \Delta P_{\max}^i]$ MW/min
-
-\textbf{Constraints}:
-\begin{align}
-P_{\min}^i &\leq P^i + \Delta P^i \leq P_{\max}^i \quad \text{(capacity)} \\
-|\Delta P^i| &\leq R_{\max}^i \quad \text{(ramp rate)}
-\end{align}
-
-\textbf{Dynamics}: Governed by swing equation
-\begin{equation}
-2H \frac{df_k}{dt} = P_{\text{gen},k} - P_{\text{load},k} - \sum_{l} \frac{D_{kl}(f_k - f_l)}{X_l}
-\end{equation}
-plus stochastic disturbances (load, renewable forecast errors, N-1 outages). Dynamics are \textbf{unknown} to agents.
-
-\textbf{Shared reward}:
-\begin{equation}
-R(s, \mathbf{a}) = -\lambda_{\text{freq}} \sum_k (f_k - 60)^2 - \lambda_{\text{cost}} \sum_i C_i |\Delta P^i| - \lambda_{\text{wear}} \sum_i W_i(|\Delta P^i|)
-\end{equation}
-with $\lambda_{\text{freq}} = 1000$, $\lambda_{\text{cost}} = 1$, $\lambda_{\text{wear}} = 0.1$, and large penalty for violations $|f_k - 60| > 0.5$ Hz.
-
-\textbf{Objective}: Find policies $\{\pi^i\}_{i=1}^N$ that maximize
-\begin{equation}
-J = \mathbb{E}\left[\sum_{t=0}^\infty \gamma^t R(s_t, \mathbf{a}_t)\right]
-\end{equation}
-subject to safety: $\Pr[|f_k - 60| > 0.5] < 0.01$ for all $k$.
-
-Discount: $\gamma = 0.95$ (20-minute horizon).
-
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-%% sections/method.tex
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-\section{Proposed Methods}
-\label{sec:method}
-
-\subsection{Multi-Agent DDPG (MADDPG)}
-MADDPG \cite{lowe2017} uses centralized training with decentralized execution.
-
-\textbf{Decentralized actors}: Each agent has policy $\pi^i(o^i; \theta^i): \mathcal{O}^i \to \mathcal{A}^i$
-
-\textbf{Centralized critics}: $Q^i(s, a^1, \ldots, a^N; \phi^i)$ uses global state and all actions during training
-
-\textbf{Training}: 
-\begin{itemize}
-    \item Critic minimizes TD error: $\mathcal{L}(\phi^i) = \mathbb{E}[(Q^i(s, \mathbf{a}) - (R + \gamma Q^i(s', \mathbf{a}')))^2]$
-    \item Actor maximizes Q via gradient: $\nabla_{\theta^i} J = \mathbb{E}[\nabla_{\theta^i} \pi^i \nabla_{a^i} Q^i|_{a^i=\pi^i}]$
-\end{itemize}
-
-\textbf{Execution}: Only actors deployed using local observations (no communication needed)
-
-\subsection{QMIX}
-QMIX \cite{rashid2018} learns factorized Q-functions. We adapt to continuous actions via NAF:
-\begin{equation}
-Q^i(o^i, a^i) = V^i(o^i) - \frac{1}{2}(a^i - \mu^i(o^i))^2 P^i(o^i)
-\end{equation}
-
-Mixing network combines local Q-values monotonically:
-\begin{equation}
-Q_{\text{tot}}(s, \mathbf{a}) = g(Q^1, \ldots, Q^N, s)
-\end{equation}
-with $\partial Q_{\text{tot}}/\partial Q^i \geq 0$ ensuring individual-global-max property.
-
-\subsection{Independent DDPG (IDDPG)}
-Baseline where each agent learns independently. Expected to underperform due to non-stationarity.
-
-\subsection{Safety Mechanisms}
-
-\textbf{Action projection}: Project policy output onto constraint set
-\begin{equation}
-a^i = \arg\min_{a \in \mathcal{C}^i} \|\tilde{a}^i - a\|^2
-\end{equation}
-where $\mathcal{C}^i$ encodes capacity and ramp constraints.
-
-\textbf{Safety critic}: Learn $Q_{\text{safe}}^i(o^i, a^i)$ predicting violation probability. Bias policy toward safe regions during execution.
-
-\textbf{Lagrangian}: For soft constraints, learn dual variables:
-\begin{equation}
-\mathcal{L} = J(\theta) + \lambda(\mathbb{E}[\text{violation}] - \delta)
-\end{equation}
-
-All methods use these safety layers.
-
-\subsection{Algorithm}
-\begin{algorithm}[H]
-\caption{MADDPG with Safety}
-\For{episode $= 1, 2, \ldots$}{
-    Reset environment, get observations $\{o_0^i\}$\;
-    \For{$t = 0, 1, \ldots, T-1$}{
-        Each agent: $\tilde{a}_t^i = \pi^i(o_t^i) + \text{noise}$\;
-        Project: $a_t^i = \text{proj}_{\mathcal{C}^i}(\tilde{a}_t^i)$\;
-        Execute $\mathbf{a}_t$, observe $R_t, s_{t+1}, \{o_{t+1}^i\}$\;
-        Store transition in replay buffer\;
-    }
-    Sample minibatch, update critics and actors\;
-    Soft update target networks\;
-}
-\end{algorithm}
-
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-%% sections/experiments.tex
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-\section{Experimental Evaluation}
-\label{sec:experiments}
-
-\subsection{Environment}
-\textbf{System}: IEEE 68-bus transmission network \\
-\textbf{Simulator}: Pandapower + custom swing dynamics \\
-\textbf{Data}: ERCOT historical data (6 months), NREL renewable profiles \\
-\textbf{Episode}: 1440 steps (24 hours at 1-min resolution)
-
-\subsection{Baselines}
-\begin{itemize}
-    \item \textbf{B1: PI-AGC}: Industry standard proportional-integral controller
-    \item \textbf{B2: MPC}: Model predictive control (oracle with perfect model)
-    \item \textbf{B3: Behavioral cloning}: Supervised learning on historical data
-\end{itemize}
-
-\subsection{Proposed Methods}
-\begin{itemize}
-    \item \textbf{M1: MADDPG}: Centralized training, decentralized execution
-    \item \textbf{M2: QMIX}: Value factorization with NAF
-    \item \textbf{M3: IDDPG}: Independent learners (baseline)
-\end{itemize}
-
-\subsection{Training}
-\begin{itemize}
-    \item 5M steps (≈3,500 episodes) across 32 parallel environments
-    \item Learning rates: $\alpha_\pi = 10^{-4}$, $\alpha_Q = 10^{-3}$
-    \item Batch size: 256, Replay buffer: 500K
-    \item 10 random seeds, report mean $\pm$ 95\% CI
-\end{itemize}
-
-\subsection{Evaluation Scenarios}
-\begin{itemize}
-    \item \textbf{S1}: Normal operation (100 episodes)
-    \item \textbf{S2}: N-1 contingencies (50 episodes) - generator/line outages
-    \item \textbf{S3}: Renewable ramps (30 episodes) - extreme forecast errors
-    \item \textbf{S4}: Distribution shift (50 episodes) - different season
-\end{itemize}
-
-\subsection{Metrics}
-\textbf{Primary}:
-\begin{itemize}
-    \item Cumulative reward $\sum_t R_t$
-    \item Frequency stability: \% time with $|f - 60| < 0.2$ Hz (target: >99\%)
-    \item Regulation cost (total \$)
-    \item Constraint violations: \% time $|f - 60| > 0.5$ Hz (target: <1\%)
-\end{itemize}
-
-\textbf{Secondary}: Area control error, max deviation, renewable utilization
-
-\subsection{Ablations}
-\begin{itemize}
-    \item Coordination: MADDPG vs. QMIX vs. IDDPG
-    \item Observations: Full state vs. local+broadcast vs. local only
-    \item Safety: No safety vs. projection vs. critic vs. both
-    \item Reward weights: Vary $\lambda_{\text{freq}}, \lambda_{\text{cost}}$
-\end{itemize}
-
-\subsection{Success Criteria}
-\textbf{Tier 1 (minimum)}: MADDPG achieves >99\% frequency stability, >15\% cost reduction, zero critical failures
-
-\textbf{Tier 2 (expected)}: MADDPG/QMIX outperform IDDPG by >15\%, handle N-1 contingencies, >25\% cost reduction
-
-\textbf{Tier 3 (stretch)}: Achieve 90\% of MPC performance, generalize to larger grids
-
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-%% sections/conclusion.tex
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-\section{Conclusion}
-\label{sec:conclusion}
-
-This project provides the first systematic comparison of MARL algorithms for power grid frequency regulation with safety constraints. We expect to demonstrate that coordinated multi-agent RL (MADDPG/QMIX) can achieve significant cost savings (25\%+) over traditional control while maintaining reliability and handling contingencies.
-
-\textbf{Future work}: Scale to larger grids (100+ buses), sim-to-real transfer, adversarial robustness, integration with electricity markets.
-
-\textbf{Broader impacts}: Enabling higher renewable penetration reduces carbon emissions and costs. However, deployment requires careful consideration of cybersecurity, accountability, and equitable access.
-
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-%% sections/app-proof-convergence.tex
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-\section{Hyperparameters and Implementation}
-
-\subsection{Network Architectures}
-\textbf{Actor}: 3-layer MLP [15, 256, 128, 64, 1] with ReLU, tanh output \\
-\textbf{Critic}: 4-layer MLP [160, 512, 256, 128, 1] with ReLU
-
-\subsection{Key Hyperparameters}
-\begin{itemize}
-    \item Discount: $\gamma = 0.95$
-    \item Target update: $\tau = 0.01$
-    \item Exploration: OU noise $\sigma = 0.2 \to 0.05$ (linear decay)
-    \item Gradient clipping: norm = 0.5
-\end{itemize}
-
-\subsection{Computational Requirements}
-Training: ≈200 GPU-hours per method on 4 NVIDIA A100s \\
-Total: ≈1000 GPU-hours (feasible with institutional cluster)
-
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-%% refs.bib (create this file)
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
-% Sample references - replace with actual citations
-
-@techreport{nerc2023,
-  author = {NERC},
-  title = {Frequency Response Initiative Report},
-  institution = {North American Electric Reliability Corporation},
-  year = {2023}
-}
-
-@book{kundur1994,
-  author = {Kundur, P.},
-  title = {Power System Stability and Control},
-  publisher = {McGraw-Hill},
-  year = {1994}
-}
-
-@article{mohamed2012,
-  author = {Mohamed, T. H. and others},
-  title = {Decentralized model predictive based load frequency control},
-  journal = {Energy Conversion and Management},
-  year = {2012}
-}
-
-@article{venkat2008,
-  author = {Venkat, A. N. and others},
-  title = {Distributed MPC strategies for automatic generation control},
-  journal = {IEEE Transactions on Control Systems Technology},
-  year = {2008}
-}
-
-@article{zhang2020,
-  author = {Zhang, Y. and others},
-  title = {Deep reinforcement learning based volt-VAR optimization},
-  journal = {IEEE Transactions on Smart Grid},
-  year = {2020}
-}
-
-@article{cao2020,
-  author = {Cao, D. and others},
-  title = {Reinforcement learning and its applications in power systems},
-  journal = {Journal of Modern Power Systems and Clean Energy},
-  year = {2020}
-}
-
-@inproceedings{achiam2017,
-  author = {Achiam, J. and others},
-  title = {Constrained Policy Optimization},
-  booktitle = {ICML},
-  year = {2017}
-}
-
-@article{dalal2018,
-  author = {Dalal, G. and others},
-  title = {Safe exploration in continuous action spaces},
-  journal = {arXiv:1801.08757},
-  year = {2018}
-}
-
-@inproceedings{alshiekh2018,
-  author = {Alshiekh, M. and others},
-  title = {Safe reinforcement learning via shielding},
-  booktitle = {AAAI},
-  year = {2018}
-}
-
-@inproceedings{tan1993,
-  author = {Tan, M.},
-  title = {Multi-agent reinforcement learning},
-  booktitle = {ICML},
-  year = {1993}
-}
-
-@inproceedings{lowe2017,
-  author = {Lowe, R. and others},
-  title = {Multi-agent actor-critic for mixed cooperative-competitive environments},
-  booktitle = {NeurIPS},
-  year = {2017}
-}
-
-@inproceedings{rashid2018,
-  author = {Rashid, T. and others},
-  title = {QMIX: Monotonic value function factorisation},
-  booktitle = {ICML},
-  year = {2018}
-}

From aa3f1ef38685a592a4b2824a11f039ff0a9b6cab Mon Sep 17 00:00:00 2001
From: Derek Smith <60520818+Br3ady@users.noreply.github.com>
Date: Sat, 1 Nov 2025 15:03:49 -0400
Subject: [PATCH 10/17] Delete preamble_packages.tex

---
 preamble_packages.tex | 135 ------------------------------------------
 1 file changed, 135 deletions(-)
 delete mode 100644 preamble_packages.tex

diff --git a/preamble_packages.tex b/preamble_packages.tex
deleted file mode 100644
index 596909b..0000000
--- a/preamble_packages.tex
+++ /dev/null
@@ -1,135 +0,0 @@
-%!TEX root = main.tex
-
-% LC: debug
-%==========================================================================
-%\usepackage{refcheck}
-%\usepackage[notref]{showkeys}
-%==========================================================================
-
-% LC: to be used for TRO
-%==========================================================================
-% \usepackage{mathptmx} % assumes new font selection scheme installed
-%\usepackage{times} % assumes new font selection scheme installed
-%==========================================================================
-\usepackage{comment}
-\usepackage{siunitx}
-\usepackage{relsize}
-\usepackage{ifthen}
-\usepackage[colorinlistoftodos]{todonotes}
-
-\usepackage[caption=false]{subfig}
-
-% \begin{comment}
-% % Fancy formatting 
-% \usepackage[tracking=false,kerning=true,spacing=true]{microtype}
-% \usepackage[caption=false]{subfig}
-
-% \usepackage[style=numeric-comp,sorting=none,firstinits=true, maxnames=3,bibstyle=numeric,abbreviate=false,defernums=true,eprint=false,backend=bibtex]{biblatex}
-% \renewcommand{\bibfont}{\footnotesize}
-% % \DeclareFieldFormat[article]{pages}{#1}%
-% % \DeclareFieldFormat[inproceedings]{pages}{#1}%
-
-% \AtEveryBibitem{%
-%   \ifentrytype{article}{%
-%     \clearfield{pages}%
-%   }{%
-%   }%
-%   \ifentrytype{inproceedings}{%
-%     \clearfield{pages}%
-%   }{%
-%   }%
-% }
-% \end{comment}
-
-% \bibliography{commons/bib/strings_long}
-% \bibliography{commons/bib/all_only_one_w_pages}
-% \bibliography{../../references/references}
-
-% \usepackage[noadjust]{cite}
-\usepackage[vlined,ruled,linesnumbered]{algorithm2e}
-\usepackage{graphics} % for pdf, bitmapped graphics files
-\usepackage{rotating}
-\usepackage{color}
-\usepackage{enumerate}
-\usepackage[T1]{fontenc}
-\usepackage{psfrag}
-\usepackage{epsfig} % for postscript graphics files
-%\usepackage{subfigure}
-% \usepackage{hyperref}
-\usepackage{booktabs}
-\usepackage{graphicx,url}
-\usepackage{multirow}
-\usepackage{array}
-\usepackage{latexsym}
-\usepackage{amsfonts}
-\usepackage{amsmath}
-\usepackage{amssymb}
-\usepackage{mathtools}
-%\usepackage{amsthm}
-\usepackage{xstring}
-\usepackage[noend]{algorithmic}
-\usepackage{multirow}
-\usepackage{xcolor}
-\usepackage{prettyref}
-\usepackage{flexisym}
-\usepackage{bigdelim}
-\usepackage{breqn} % load this last
-\usepackage{listings}
-\let\labelindent\relax
-\usepackage{enumitem}
-\usepackage{xspace}
-\usepackage{bm}
-\graphicspath{{./figures/}}
-\usepackage{tikz}
-\usetikzlibrary{matrix,calc}
-
-
-%\usepackage{ifpdf}
-% Heiko Oberdiek's ifpdf.sty is very useful if you need conditional
-% compilation based on whether the output is pdf or dvi.
-% usage:
-% \ifpdf
-%   % pdf code
-% \else
-%    dvi code
-% \fi
-% The latest version of ifpdf.sty can be obtained from:
-% http://www.ctan.org/tex-archive/macros/latex/contrib/oberdiek/
-% Also, note that IEEEtran.cls V1.7 and later provides a builtin
-% \ifCLASSINFOpdf conditional that works the same way.
-% When switching from latex to pdflatex and vice-versa, the compiler may
-% have to be run twice to clear warning/error messages.
-
-% *** GRAPHICS RELATED PACKAGES ***
-%
-%\ifCLASSINFOpdf
-  % \usepackage[pdftex]{graphicx}
-  % declare the path(s) where your graphic files are
-  % \graphicspath{{../pdf/}{../jpeg/}}
-  % and their extensions so you won't have to specify these with
-  % every instance of \includegraphics
-  % \DeclareGraphicsExtensions{.pdf,.jpeg,.png}
-%\else
-  % or other class option (dvipsone, dvipdf, if not using dvips). graphicx
-  % will default to the driver specified in the system graphics.cfg if no
-  % driver is specified.
-  % \usepackage[dvips]{graphicx}
-  % declare the path(s) where your graphic files are
-  % \graphicspath{{../eps/}}
-  % and their extensions so you won't have to specify these with
-  % every instance of \includegraphics
-  % \DeclareGraphicsExtensions{.eps}
-%\fi
-
-
-\usepackage{mdwlist}
-\let\stditemize\itemize
-\let\endstditemize\enditemize
-\let\itemize\undefined
-\makecompactlist{itemize}{stditemize}
-
-%\let\stdenumerate\enumerate
-%\let\endstdenumerate\endenumerate
-%\let\enumerate\undefined
-%\makecompactlist{enumerate}{stdenumerate}
- 

From 5a04115df02746067981aac55c901b38c3e1b063 Mon Sep 17 00:00:00 2001
From: Derek Smith <60520818+Br3ady@users.noreply.github.com>
Date: Sat, 1 Nov 2025 15:04:06 -0400
Subject: [PATCH 11/17] Delete preamble_symbols.tex

---
 preamble_symbols.tex | 452 -------------------------------------------
 1 file changed, 452 deletions(-)
 delete mode 100644 preamble_symbols.tex

diff --git a/preamble_symbols.tex b/preamble_symbols.tex
deleted file mode 100644
index 584697f..0000000
--- a/preamble_symbols.tex
+++ /dev/null
@@ -1,452 +0,0 @@
-%!TEX root = main.tex
-
-% \newcommand{\qed}{{\hfill $\square$}}
-
-% Format definition
-\newrefformat{prob}{Problem\,\ref{#1}}
-\newrefformat{def}{Definition\,\ref{#1}}
-\newrefformat{sec}{Section\,\ref{#1}}
-\newrefformat{sub}{Section\,\ref{#1}}
-\newrefformat{prop}{Proposition\,\ref{#1}}
-\newrefformat{app}{Appendix\,\ref{#1}}
-\newrefformat{alg}{Algorithm\,\ref{#1}}
-\newrefformat{cor}{Corollary\,\ref{#1}}
-\newrefformat{thm}{Theorem\,\ref{#1}}
-\newrefformat{lem}{Lemma\,\ref{#1}}
-\newrefformat{fig}{Fig.\,\ref{#1}}
-\newrefformat{tab}{Table\,\ref{#1}}
-
-% Problem environment
-\newtheorem{theorem}{Theorem}
-\newtheorem{problem}{Problem}
-\newtheorem{trule}[theorem]{Rule}
-\newtheorem{corollary}[theorem]{Corollary}
-%\newtheorem{algorithm}{Algorithm}
-%\newtheorem{procedure}{\textbf{Procedure}}
-\newtheorem{conjecture}[theorem]{Conjecture}
-\newtheorem{lemma}[theorem]{Lemma}
-\newtheorem{assumption}[theorem]{Assumption}
-\newtheorem{definition}[theorem]{Definition}
-\newtheorem{proposition}[theorem]{Proposition}
-\newtheorem{remark}[theorem]{Remark}
-\newtheorem{example}[theorem]{Example}
-
-% Shortcuts
-\newcommand{\cf}{\emph{cf.}\xspace}
-\newcommand{\Figure}{Fig.}
-\newcommand{\bdmath}{\begin{dmath}}
-\newcommand{\edmath}{\end{dmath}}
-\newcommand{\beq}{\begin{equation}}
-\newcommand{\eeq}{\end{equation}}
-\newcommand{\bdm}{\begin{displaymath}}
-\newcommand{\edm}{\end{displaymath}}
-\newcommand{\bea}{\begin{eqnarray}}
-\newcommand{\eea}{\end{eqnarray}}
-\newcommand{\beal}{\beq \begin{array}{ll}}
-\newcommand{\eeal}{\end{array} \eeq}
-\newcommand{\beas}{\begin{eqnarray*}}
-\newcommand{\eeas}{\end{eqnarray*}}
-\newcommand{\ba}{\begin{array}}
-\newcommand{\ea}{\end{array}}
-\newcommand{\bit}{\begin{itemize}}
-\newcommand{\eit}{\end{itemize}}
-\newcommand{\ben}{\begin{enumerate}}
-\newcommand{\een}{\end{enumerate}}
-\newcommand{\expl}[1]{&&\qquad\text{\color{gray}(#1)}\nonumber}
-
-% \newcommand{\insertproof}[1]{%
-%     % lem:name -> proof_lem_name.tex
-%     \StrSubstitute{#1}{:}{_}[\name] % replace : with _
-%     \begin{IEEEproof} %\color[rgb]{0.8,0.8,0.8}
-%     \input{proof_\name.tex}
-%     \end{IEEEproof}
-% }
-
-% Calligraphic fonts
-\newcommand{\calA}{{\cal A}}
-\newcommand{\calB}{{\cal B}}
-\newcommand{\calC}{{\cal C}}
-\newcommand{\calD}{{\cal D}}
-\newcommand{\calE}{{\cal E}}
-\newcommand{\calF}{{\cal F}}
-\newcommand{\calG}{{\cal G}}
-\newcommand{\calH}{{\cal H}}
-\newcommand{\calI}{{\cal I}}
-\newcommand{\calJ}{{\cal J}}
-\newcommand{\calK}{{\cal K}}
-\newcommand{\calL}{{\cal L}}
-\newcommand{\calM}{{\cal M}}
-\newcommand{\calN}{{\cal N}}
-\newcommand{\calO}{{\cal O}}
-\newcommand{\calP}{{\cal P}}
-\newcommand{\calQ}{{\cal Q}}
-\newcommand{\calR}{{\cal R}}
-\newcommand{\calS}{{\cal S}}
-\newcommand{\calT}{{\cal T}}
-\newcommand{\calU}{{\cal U}}
-\newcommand{\calV}{{\cal V}}
-\newcommand{\calW}{{\cal W}}
-\newcommand{\calX}{{\cal X}}
-\newcommand{\calY}{{\cal Y}}
-\newcommand{\calZ}{{\cal Z}}
-
-% SETS:
-\newcommand{\setA}{\textsf{A}}
-\newcommand{\setB}{\textsf{B}}
-\newcommand{\setC}{\textsf{C}}
-\newcommand{\setD}{\textsf{D}}
-\newcommand{\setE}{\textsf{E}}
-\newcommand{\setF}{\textsf{F}}
-\newcommand{\setG}{\textsf{G}}
-\newcommand{\setH}{\textsf{H}}
-\newcommand{\setI}{\textsf{I}}
-\newcommand{\setJ}{\textsf{J}}
-\newcommand{\setK}{\textsf{K}}
-\newcommand{\setL}{\textsf{L}}
-\newcommand{\setM}{\textsf{M}}
-\newcommand{\setN}{\textsf{N}}
-\newcommand{\setO}{\textsf{O}}
-\newcommand{\setP}{\textsf{P}}
-\newcommand{\setQ}{\textsf{Q}}
-\newcommand{\setR}{\textsf{R}}
-\newcommand{\setS}{\textsf{S}}
-\newcommand{\setT}{\textsf{T}}
-\newcommand{\setU}{\textsf{U}}
-\newcommand{\setV}{\textsf{V}}
-\newcommand{\setX}{\textsf{X}}
-\newcommand{\setY}{\textsf{Y}}
-\newcommand{\setZ}{\textsf{Z}}
-
-%General
-\newcommand{\smallheading}[1]{\textit{#1}: }
-\newcommand{\algostep}[1]{{\small\texttt{#1:}}\xspace}
-\newcommand{\etal}{\emph{et~al.}\xspace}
-\newcommand{\setal}{~\emph{et~al.}\xspace}
-\newcommand{\eg}{\emph{e.g.,}\xspace}
-\newcommand{\ie}{\emph{i.e.,}\xspace}
-\newcommand{\myParagraph}[1]{{\bf #1.}\xspace}
-% \newcommand{\email}[1]{{\smaller \textsf{#1}}}
-
-%Typography
-\newcommand{\M}[1]{{\bm #1}} % Face for matrices
-\renewcommand{\boldsymbol}[1]{{\bm #1}}
-\newcommand{\algoname}[1]{\textsc{#1}} % Name of algorithms
-
-%Editing
-\newcommand{\towrite}{{\color{red} To write}\xspace}
-\newcommand{\xxx}{{\color{red} XXX}\xspace}
-\newcommand{\AC}[1]{{\color{red} \textbf{AC}: #1}}
-\newcommand{\LC}[1]{{\color{red} \textbf{LC}: #1}}
-\newcommand{\FM}[1]{{\color{red} \textbf{FM}: #1}}
-\newcommand{\RT}[1]{{\color{blue} \textbf{RT}: #1}}
-\newcommand{\VT}[1]{{\color{blue} \textbf{VT}: #1}}
-\newcommand{\hide}[1]{}
-\newcommand{\wrt}{w.r.t.\xspace}
-\newcommand{\highlight}[1]{{\color{red} #1}}
-\newcommand{\tocheck}[1]{{\color{brown} #1}}
-\newcommand{\grayout}[1]{{\color{gray} #1}}
-\newcommand{\grayText}[1]{{\color{gray} \text{#1} }}
-\newcommand{\grayMath}[1]{{\color{gray} #1 }}
-\newcommand{\hiddenText}{{\color{gray} hidden text.}}
-\newcommand{\hideWithText}[1]{\hiddenText}
-
-
-\newcommand{\NA}{{\sf n/a}}
-\newcommand{\versus}{\scenario{VS}\xspace}
-
-%Basic math symbols
-\newcommand{\kron}{\otimes}
-\newcommand{\dist}{\mathbf{dist}}
-\newcommand{\iter}{\! \rm{iter.} \;}
-\newcommand{\leqt}{\!\!\! < \!\!\!}
-\newcommand{\geqt}{\!\!\! > \!\!\!}
-\newcommand{\mysetminus}{-} % One set minus another
-\newcommand{\powerset}{\mathcal{P}}
-\newcommand{\Int}[1]{ { {\mathbb Z}^{#1} } }
-\newcommand{\Natural}[1]{ { {\mathbb N}^{#1} } }
-\newcommand{\Complex}[1]{ { {\mathbb C}^{#1} } }
-\newcommand{\one}{ {\mathbf{1}} }
-\newcommand{\subject}{\text{ subject to }}
-\DeclareMathOperator*{\argmax}{arg\,max}
-\DeclareMathOperator*{\argmin}{arg\,min}
-
-%% Norms
-\newcommand{\normsq}[2]{\left\|#1\right\|^2_{#2}}
-\newcommand{\norm}[1]{\left\| #1 \right\|}
-\newcommand{\normsqs}[2]{\|#1\|^2_{#2}}
-\newcommand{\infnorm}[1]{\|#1\|_{\infty}}
-\newcommand{\zeronorm}[1]{\|#1\|_{0}}
-\newcommand{\onenorm}[1]{\|#1\|_{1}}
-\newcommand{\lzero}{\ell_{0}}
-\newcommand{\lone}{\ell_{1}}
-\newcommand{\linf}{\ell_{\infty}}
-
-\newcommand{\E}{{\mathbb{E}}}
-\newcommand{\EV}{\mathbb{E}}
-\newcommand{\erf}{{\mathbf{erf}}}
-\newcommand{\prob}[1]{{\mathbb P}\left(#1\right)}
-% \newcommand{\tran}{^{\top}}
-\newcommand{\tran}{^{\mathsf{T}}}
-\newcommand{\traninv}{^{-\mathsf{T}}}
-\newcommand{\diag}[1]{\mathrm{diag}\left(#1\right)}
-\newcommand{\trace}[1]{\mathrm{tr}\left(#1\right)}
-\newcommand{\conv}[1]{\mathrm{conv}\left(#1\right)}
-\newcommand{\polar}[1]{\mathrm{polar}\left(#1\right)}
-\newcommand{\rank}[1]{\mathrm{rank}\left(#1\right)}
-\newcommand{\e}{{\mathrm e}}
-\newcommand{\inv}{^{-1}}
-\newcommand{\pinv}{^\dag}
-\newcommand{\until}[1]{\{1,\dots, #1\}}
-\newcommand{\ones}{{\mathbf 1}}
-\newcommand{\zero}{{\mathbf 0}}
-\newcommand{\eye}{{\mathbf I}}
-\newcommand{\vect}[1]{\left[\begin{array}{c}  #1  \end{array}\right]}
-\newcommand{\matTwo}[1]{\left[\begin{array}{cc}  #1  \end{array}\right]}
-\newcommand{\matThree}[1]{\left[\begin{array}{ccc}  #1  \end{array}\right]}
-\newcommand{\dss}{\displaystyle}
-\newcommand{\Real}[1]{ { {\mathbb R}^{#1} } }
-\newcommand{\reals}{\Real{}}
-\newcommand{\opt}{^{\star}}
-\newcommand{\only}{^{\alpha}}
-\newcommand{\copt}{^{\text{c}\star}}
-\newcommand{\atk}{^{(k)}}
-\newcommand{\att}{^{(t)}}
-\newcommand{\at}[1]{^{(#1)}}
-\newcommand{\attau}{^{(\tau)}}
-\newcommand{\atc}[1]{^{(#1)}}
-\newcommand{\atK}{^{(K)}}
-\newcommand{\atj}{^{(j)}}
-\newcommand{\projector}{{\tt projector}}
-\newcommand{\setdef}[2]{ \{#1 \; {:} \; #2 \} }
-\newcommand{\smalleye}{\left(\begin{smallmatrix}1&0\\0&1\end{smallmatrix}\right)}
-
-%Spaces
-\newcommand{\SEtwo}{\ensuremath{\mathrm{SE}(2)}\xspace}
-\newcommand{\SE}[1]{\ensuremath{\mathrm{SE}(#1)}\xspace}
-\newcommand{\SEthree}{\ensuremath{\mathrm{SE}(3)}\xspace}
-\newcommand{\SOtwo}{\ensuremath{\mathrm{SO}(2)}\xspace}
-\newcommand{\SOthree}{\ensuremath{\mathrm{SO}(3)}\xspace}
-\newcommand{\Othree}{\ensuremath{\mathrm{O}(3)}\xspace}
-\newcommand{\SOn}{\ensuremath{\mathrm{SO}(n)}\xspace}
-\newcommand{\SO}[1]{\ensuremath{\mathrm{SO}(#1)}\xspace}
-\newcommand{\On}{\ensuremath{\mathrm{O}(n)}\xspace}
-\newcommand{\sotwo}{\ensuremath{\mathrm{so}(2)}\xspace}
-\newcommand{\sothree}{\ensuremath{\mathrm{so}(3)}\xspace}
-\newcommand{\intexpmap}[1]{\mathrm{Exp}\left(#1\right)}
-\newcommand{\intlogmap}[1]{\mathrm{Log}\left(#1\right)}
-\newcommand{\logmapz}[1]{\mathrm{Log}_0(#1)}
-\newcommand{\loglikelihood}{\mathrm{log}\calL}
-\newcommand{\intprinlogmap}{\mathrm{Log}_0}
-\newcommand{\expmap}[1]{\intexpmap{#1}}
-\newcommand{\expmaps}[1]{\langle #1 \rangle_{2\pi}}
-\newcommand{\biggexpmap}[1]{\left\langle #1 \right\rangle_{2\pi}}
-\newcommand{\logmap}[1]{\intlogmap{#1}}
-\newcommand{\round}[1]{\mathrm{round}\left( #1 \right)}
-
-% Matrices 
-\newcommand{\MA}{\M{A}}
-\newcommand{\MB}{\M{B}}
-\newcommand{\MC}{\M{C}}
-\newcommand{\MD}{\M{D}}
-\newcommand{\ME}{\M{E}}
-\newcommand{\MJ}{\M{J}}
-\newcommand{\MK}{\M{K}}
-\newcommand{\MG}{\M{G}}
-\newcommand{\MM}{\M{M}}
-\newcommand{\MN}{\M{N}}
-\newcommand{\MP}{\M{P}}
-\newcommand{\MQ}{\M{Q}}
-\newcommand{\MU}{\M{U}}
-\newcommand{\MR}{\M{R}}
-\newcommand{\MS}{\M{S}}
-\newcommand{\MI}{\M{I}}
-\newcommand{\MV}{\M{V}}
-\newcommand{\MF}{\M{F}}
-\newcommand{\MH}{\M{H}}
-\newcommand{\ML}{\M{L}}
-\newcommand{\MO}{\M{O}}
-\newcommand{\MT}{\M{T}}
-\newcommand{\MX}{\M{X}}
-\newcommand{\MY}{\M{Y}}
-\newcommand{\MW}{\M{W}}
-\newcommand{\MZ}{\M{Z}}
-\newcommand{\MSigma}{\M{\Sigma}} 
-\newcommand{\MOmega}{\M{\Omega}}
-\newcommand{\MPhi}{\M{\Phi}}
-\newcommand{\MPsi}{\M{\Psi}}
-\newcommand{\MDelta}{\M{\Delta}}
-\newcommand{\MLambda}{\M{\Lambda}}
-
-% vectors
-\newcommand{\va}{\boldsymbol{a}} 
-\newcommand{\vh}{\boldsymbol{h}} 
-\newcommand{\vb}{\boldsymbol{b}}
-\newcommand{\vc}{\boldsymbol{c}}
-\newcommand{\vd}{\boldsymbol{d}}
-\newcommand{\ve}{\boldsymbol{e}}
-\newcommand{\vf}{\boldsymbol{f}}
-\newcommand{\vg}{\boldsymbol{g}}
-\newcommand{\vk}{\boldsymbol{k}}
-\newcommand{\vl}{\boldsymbol{l}}
-\newcommand{\vn}{\boldsymbol{n}}
-\newcommand{\vo}{\boldsymbol{o}}
-\newcommand{\vp}{\boldsymbol{p}}
-\newcommand{\vq}{\boldsymbol{q}}
-\newcommand{\vr}{\boldsymbol{r}}
-\newcommand{\vs}{\boldsymbol{s}}
-\newcommand{\vu}{\boldsymbol{u}}
-\newcommand{\vv}{\boldsymbol{v}}
-\newcommand{\vt}{\boldsymbol{t}}
-\newcommand{\vxx}{\boldsymbol{x}} 
-\newcommand{\vy}{\boldsymbol{y}}
-\newcommand{\vw}{\boldsymbol{w}}
-\newcommand{\vzz}{\boldsymbol{z}} 
-\newcommand{\vdelta}{\boldsymbol{\delta}}
-\newcommand{\vgamma}{\boldsymbol{\gamma}}  
-\newcommand{\vlambda}{\boldsymbol{\lambda}}
-\newcommand{\vtheta}{\boldsymbol{\theta}}
-\newcommand{\valpha}{\boldsymbol{\alpha}}
-\newcommand{\vbeta}{\boldsymbol{\beta}}
-\newcommand{\vnu}{\boldsymbol{\nu}}
-\newcommand{\vmu}{\boldsymbol{\mu}}
-\newcommand{\vepsilon}{\boldsymbol{\epsilon}}
-\newcommand{\vtau}{\boldsymbol{\tau}}
-
-%Intrinsic geometry
-\newcommand{\Rtheta}{\boldsymbol{R}}
-\newcommand{\symf}{f} % Symmetry function
-
-%Angles
-\newcommand{\angledomain}{(-\pi,+\pi]}
-
-% Tree, graphs, and cycle basis
-\newcommand{\MCB}{\mathsf{MCB}}
-\newcommand{\FCM}{\mathsf{FCM}}
-\newcommand{\FCB}{\mathsf{FCB}}
-\newcommand{\cyclemap}[1]{\calC^{\calG}\left(#1\right)}
-\newcommand{\incidencemap}[1]{\calA^{\calG}\left(#1\right)}
-\newcommand{\cyclemapk}{\calC^{\calG}_{k}}
-\newcommand{\incidencemapij}{\calA^{\calG}_{ij}}
-\renewcommand{\ij}{_{ij}}
-\newcommand{\foralledges}{\forall(i,j) \in \calE}
-\newcommand{\sumalledges}{
-     \displaystyle
-     \sum_{(i,j) \in \calE}}
-\newcommand{\sumalledgesm}{
-     \displaystyle
-     \sum_{i=1}^{m}}
-\newcommand{\T}{\mathsf{T}}
-\newcommand{\To}{\T_{\rm o}}
-\newcommand{\Tm}{\T_{\rm m}}
-\newcommand{\MCBa}{\MCB_{\mathsf{a}}}
-\newcommand{\FCBo}{\FCB_{\mathsf{o}}}
-\newcommand{\FCBm}{\FCB_{\mathsf{m}}}
-
-% Algorithms
-\newcommand{\algoonlyname}{MOLE2D}
-\newcommand{\algoml}{{\smaller\sf \algoonlyname}\xspace}
-\newcommand{\algocyclebasis}{\algoname{compute-cycle-basis}}
-\newcommand{\scenario}[1]{{\smaller \sf#1}\xspace}
-\newcommand{\toro}{{\smaller\sf Toro}\xspace}
-\newcommand{\gtwoo}{{\smaller\sf g2o}\xspace}
-\newcommand{\gtwooST}{{\smaller\sf g2oST}\xspace}
-\newcommand{\gtwood}{{\smaller\sf g2o{10}}\xspace}
-\newcommand{\gtsam}{{\smaller\sf gtsam}\xspace}
-\newcommand{\isam}{{\smaller\sf iSAM}\xspace}
-\newcommand{\lago}{{\smaller\sf LAGO}\xspace}
-\newcommand{\egtwoo}{{\smaller\sf \algoonlyname+g2o}\xspace}
-
-% Datasets
-%\newcommand{\grid}{\scenario{cube}}
-\newcommand{\rim}{\scenario{rim}}
-\newcommand{\cubicle}{\scenario{cubicle}}
-\newcommand{\sphere}{\scenario{sphere}}
-\newcommand{\sphereHard}{\scenario{sphere-a}}
-\newcommand{\garage}{\scenario{garage}}
-\newcommand{\torus}{\scenario{torus}}
-\newcommand{\oneloop}{\scenario{circle}}
-\newcommand{\intel}{\scenario{INTEL}}
-\newcommand{\bovisa}{\scenario{Bovisa}}
-\newcommand{\bov}{\scenario{B25b}}
-\newcommand{\fra}{\scenario{FR079}}
-\newcommand{\frb}{\scenario{FRH}}
-\newcommand{\csail}{\scenario{CSAIL}}
-\newcommand{\Ma}{\scenario{M3500}}
-\newcommand{\Mb}{\scenario{M10000}}
-\newcommand{\ATE}{\scenario{ATE}}
-\newcommand{\CVX}{\scenario{CVX}}
-\newcommand{\NEOS}{\scenario{NEOS}}
-\newcommand{\sdptThree}{\scenario{sdpt3}}
-\newcommand{\MOSEK}{\scenario{MOSEK}}
-\newcommand{\NESTA}{\scenario{NESTA}}
-\newcommand{\vertigo}{\scenario{Vertigo}}
-\newcommand{\SDPA}{\scenario{SDPA}}
-
-% \newcommand{\tablabel}[1]{%
-%     \hspace{-1mm}%
-%     \begin{sideways}{\small\scenario{#1}}\end{sideways}%
-%     \hspace{-4mm}%
-% }
-%
-% \newcommand{\tabfig}[2]{%
-%     \subfloat[\label{fig:#2} #1]{%
-%     \begin{minipage}{5.5cm}%
-%     \centering%
-%     \includegraphics[width=5.5cm,trim=4 0 4 15]{figures/#2}
-%     \end{minipage}%
-%     }%
-% }
-%
-% \newcommand{\subFigure}[3]{%
-%     \subfloat[\label{fig:#2} #1]{%
-%     \begin{minipage}{#3cm}%
-%     \centering%
-%     \includegraphics[width=#3cm,trim=4 0 4 15]{figures/#2}
-%     \end{minipage}%
-%     }%
-% }
-\newcommand{\cvx}{{\sf cvx}\xspace}
-
-% COLORS
-\newcommand{\blue}[1]{{\color{blue}#1}}
-\newcommand{\green}[1]{{\color{green}#1}}
-\newcommand{\red}[1]{{\color{red}#1}}
-
-% TO MANAGE REFERENCES
-%============================================================================
-\newcommand{\linkToPdf}[1]{\href{#1}{\blue{(pdf)}}}
-\newcommand{\linkToPpt}[1]{\href{#1}{\blue{(ppt)}}}
-\newcommand{\linkToCode}[1]{\href{#1}{\blue{(code)}}}
-\newcommand{\linkToWeb}[1]{\href{#1}{\blue{(web)}}}
-\newcommand{\linkToVideo}[1]{\href{#1}{\blue{(video)}}}
-\newcommand{\linkToMedia}[1]{\href{#1}{\blue{(media)}}}
-\newcommand{\award}[1]{\xspace} % {{\red{#1}}} % omit awards
-
-
-% PAPER-SPECIFIC COMMANDS
-%============================================================================
-
-% Linear approximation
-\newcommand{\vpose}{\boldsymbol{x}}
-\newcommand{\vz}{\boldsymbol{z}}
-\newcommand{\vDelta}{\boldsymbol{\Delta}}
-\newcommand{\vposesub}{\hat{\vpose}}
-\newcommand{\vpossub}{\hat{\vpos}}
-\newcommand{\vthetasub}{\hat{\vtheta}}
-\newcommand{\Pthetasub}{\MP_\vtheta}
-\newcommand{\thetasub}{\hat{\theta}}
-\newcommand{\vposecorr}{\tilde{\vpose}}
-\newcommand{\vposcorr}{\tilde{\vpos}}
-\newcommand{\vthetacorr}{\tilde{\vtheta}}
-\newcommand{\vposestar}{{\vpose}^{\star}}
-\newcommand{\vposstar}{{\vpos}^{\star}}
-\newcommand{\vthetastar}{{\vtheta}^{\star}}
-\newcommand{\vcthetastar}{{\vctheta}^{\star}}
-\newcommand{\pose}{\boldsymbol{x}}
-\newcommand{\pos}{\boldsymbol{p}}
-\newcommand{\mease}{z} % element
-\newcommand{\meas}{\boldsymbol{z}} % vector
-\newcommand{\meashat}{\hat{\meas}}
-
-
-

From 0db6c763326b8d4dccc6c6e54f59af513a6a0e8e Mon Sep 17 00:00:00 2001
From: Derek Smith <60520818+Br3ady@users.noreply.github.com>
Date: Sat, 1 Nov 2025 15:04:19 -0400
Subject: [PATCH 12/17] Delete shortcuts.tex

---
 shortcuts.tex | 106 --------------------------------------------------
 1 file changed, 106 deletions(-)
 delete mode 100644 shortcuts.tex

diff --git a/shortcuts.tex b/shortcuts.tex
deleted file mode 100644
index 9014fd8..0000000
--- a/shortcuts.tex
+++ /dev/null
@@ -1,106 +0,0 @@
-%!TEX root = main.tex
-
-\newcommand{\Sn}{\mathbb{S}^n}
-\newcommand{\R}{\mathbb{R}}
-\newcommand{\cA}{\mathcal{A}}
-\newcommand{\cB}{\mathcal{B}}
-\newcommand{\cL}{\mathcal{L}}
-\renewcommand{\norm}[1]{\left\lVert #1 \right\rVert}
-\newcommand{\inprod}[2]{\left\langle #1, #2 \right\rangle}
-\newcommand{\vectorize}[1]{\mathrm{vec}\parentheses{#1}}
-
-\newcommand{\nnReal}[1]{\mathbb{R}_{+}^{#1}}
-\newcommand{\vcat}{\ ;\ }
-\newcommand{\mymid}{\ \middle\vert\ }
-\newcommand{\cbrace}[1]{\left\{#1\right\}}
-\newcommand{\sym}[1]{\mathbb{S}^{#1}}
-\newcommand{\calAadj}{\calA^{*}}
-\newcommand{\bary}{\bar{y}}
-\newcommand{\barC}{\bar{C}}
-\newcommand{\barcalA}{\bar{\calA}}
-\newcommand{\barcalAadj}{\bar{\calA}^{*}}
-\newcommand{\bmat}{\left[ \begin{array}}
-\newcommand{\emat}{\end{array}\right]}
-\newcommand{\psd}[1]{\sym{#1}_{+}}
-\newcommand{\parentheses}[1]{\left(#1\right)}
-\newcommand{\half}{\frac{1}{2}}
-
-\newcommand{\usphere}[1]{\calS^{#1}}
-\newcommand{\hpartial}{\hat{\partial}}
-\newcommand{\baralpha}{\bar{\alpha}}
-\newcommand{\tldW}{\widetilde{\MW}}
-\newcommand{\bareta}{\bar{\eta}}
-\newcommand{\tWnu}{\tilde{W}_{\nu}}
-\newcommand{\tWzero}{\tilde{W}_{0}}
-\newcommand{\tWone}{\tilde{W}_{1}}
-\newcommand{\tWhalf}{\tilde{W}_{1/2}}
-\newcommand{\Qa}{Q_{\alpha}}
-\newcommand{\Qb}{Q_{\baralpha}}
-\newcommand{\tV}{\tilde{V}}
-\newcommand{\tVzero}{\tV_{0}}
-\newcommand{\tVhalf}{\tV_{1/2}}
-\newcommand{\tVone}{\tV_{1}}
-\newcommand{\tVnu}{\tV_{\nu}}
-\newcommand{\Mnu}{M_{\nu}}
-\newcommand{\tMnu}{\tilde{M}_{\nu}}
-\newcommand{\tM}{\tilde{M}}
-\newcommand{\hV}{\hat{V}}
-\newcommand{\hM}{\hat{M}}
-\newcommand{\hMnu}{\hat{M}_{\nu}}
-\newcommand{\Omegahalf}{\Omega^{.5}}
-\newcommand{\bracket}[1]{\left[#1\right]}
-\newcommand{\abs}[1]{\left|#1\right|}
-\newcommand{\dx}{\dot{x}}
-\newcommand{\bbU}{\mathbb{U}}
-\newcommand{\bbX}{\mathbb{X}}
-\newcommand{\eqcon}[1]{c_{\mathrm{eq},#1}}
-\newcommand{\ineqcon}[1]{c_{\mathrm{ineq},#1}}
-\newcommand{\epsfeqcon}[1]{c_{f,\mathrm{eq},#1}}
-\newcommand{\epsfineqcon}[1]{c_{f,\mathrm{ineq},#1}}
-\newcommand{\epsgeqcon}[1]{c_{g,\mathrm{eq},#1}}
-\newcommand{\epsgineqcon}[1]{c_{g,\mathrm{ineq},#1}}
-\newcommand{\numeqepsf}{l_{f,\mathrm{eq}}}
-\newcommand{\numineqepsf}{l_{f,\mathrm{ineq}}}
-\newcommand{\numeqepsg}{l_{g,\mathrm{eq}}}
-\newcommand{\numineqepsg}{l_{g,\mathrm{ineq}}}
-\newcommand{\numeq}{l_{\mathrm{eq}}}
-\newcommand{\numineq}{l_{\mathrm{ineq}}}
-\newcommand{\epsf}{\epsilon_f}
-\newcommand{\epsg}{\epsilon_g}
-\newcommand{\longmid}{\ \middle\vert\ }
-\newcommand{\setxepsf}{\mathbb{X}_{f}}
-\newcommand{\setxepsg}{\mathbb{X}_{g}}
-\newcommand{\dotb}{\dot{b}}
-\newcommand{\poly}[1]{\mathbb{R}[#1]}
-\newcommand{\sos}[1]{\Sigma[#1]}
-\newcommand{\peps}{p_{\epsilon}}
-\newcommand{\qeps}{q_{\epsilon}}
-\newcommand{\ceil}[1]{\left\lceil #1 \right\rceil}
-\newcommand{\Ceq}[1]{C_{\mathrm{eq},#1}}
-\newcommand{\Cineq}[1]{C_{\mathrm{ineq},#1}}
-\newcommand{\uineqcon}[1]{c_{u,#1}}
-\newcommand{\numinequ}{l_{u}}
-\newcommand{\bbK}{\mathbb{K}}
-\newcommand{\numtheta}{n_{\theta}}
-\newcommand{\uref}{u_{\mathrm{ref}}}
-\newcommand{\xc}{x_c}
-\newcommand{\yc}{y_c}
-\newcommand{\xo}{x_o}
-\newcommand{\yo}{y_o}
-\newcommand{\HYedit}[1]{\blue{#1}}
-\newcommand{\tldp}{\widetilde{p}}
-\newcommand{\hatp}{\widehat{p}}
-\newcommand{\tldR}{\widetilde{R}}
-\newcommand{\hats}{\widehat{s}}
-\newcommand{\hatR}{\widehat{R}}
-\newcommand{\hatt}{\widehat{t}}
-\newcommand{\tldcalY}{\widetilde{\calY}}
-\newcommand{\hatcalR}{\widehat{\calR}}
-\newcommand{\tldcalR}{\widetilde{\calR}}
-\newcommand{\tldTheta}{\widetilde{\Theta}}
-\newcommand{\floor}[1]{\lfloor #1 \rfloor}
-\newcommand{\mysum}[1]{\text{sum}\parentheses{#1}}
-\newcommand{\barcalI}{\bar{\calI}}
-\newcommand{\barpi}{\bar{\pi}}
-\newcommand{\sosone}{\mathrm{SOS}\text{-}1}
-\newcommand{\ab}{\widetilde{ab}}
\ No newline at end of file

From 285216f9726793da4ccdcc53fedd43f42b27268f Mon Sep 17 00:00:00 2001
From: Derek Smith <60520818+Br3ady@users.noreply.github.com>
Date: Sat, 1 Nov 2025 15:07:21 -0400
Subject: [PATCH 13/17] Add files via upload

---
 proposal_3page.tex | 160 +++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 160 insertions(+)
 create mode 100644 proposal_3page.tex

diff --git a/proposal_3page.tex b/proposal_3page.tex
new file mode 100644
index 0000000..cdaaaf7
--- /dev/null
+++ b/proposal_3page.tex
@@ -0,0 +1,160 @@
+\documentclass[11pt]{article}
+\usepackage[utf8]{inputenc}
+\usepackage{amsmath}
+\usepackage{amssymb}
+\usepackage{geometry}
+\usepackage{enumitem}
+\usepackage{hyperref}
+
+\geometry{margin=1in}
+
+\title{Multi-Agent Reinforcement Learning for Real-Time Frequency Regulation in Power Grids}
+\author{Derek Smith\\
+ES 158: Sequential Decision Making in Dynamic Environments}
+\date{September 29, 2025}
+
+\begin{document}
+
+\maketitle
+
+\section{Relevance to the Course}
+
+This project addresses distributed optimal control in power grid frequency regulation, formulated as a \textbf{Multi-Agent Markov Decision Process (MA-MDP)}:
+
+\textbf{Decision-makers}: $N = 20$ controllable units (batteries, gas generators, demand response) coordinating to maintain 60 Hz frequency.
+
+\textbf{Dynamics}: Grid frequency evolves via swing equations with coupled electromechanical dynamics:
+\begin{equation}
+\frac{df}{dt} = \frac{P_{\text{gen}} - P_{\text{load}} - P_{\text{losses}}}{2H \cdot S_{\text{base}}}
+\end{equation}
+where each agent's action $\Delta P^i$ affects total $P_{\text{gen}}$.
+
+\textbf{Sequential nature}: Control decisions require multi-step lookahead due to renewable forecasts, load fluctuations, and other agents' actions. Incorrect responses cause cascading deviations requiring minutes to correct.
+
+The problem exhibits core RL challenges: continuous state/action spaces, partial observability (local measurements with delays), stochastic disturbances (renewable intermittency), and safety constraints (frequency within $\pm 0.5$ Hz). Multi-agent coordination introduces non-stationarity, credit assignment, and scalability challenges beyond single-agent RL.
+
+\section{Motivation and Related Work}
+
+\textbf{Motivation}: Renewable energy integration ($>30\%$ generation) disrupts grid operations due to lack of inertia, causing faster frequency dynamics, doubled rate-of-change-of-frequency~\cite{nerc2023}, and \$10B+ annual regulation costs. Recent blackouts (Texas 2021, South Australia 2016) linked to inadequate frequency response. Multi-agent RL offers coordinated, adaptive control potentially reducing costs 20--40\%~\cite{venkat2022}.
+
+\textbf{Prior Work}: Classical AGC uses PI controllers~\cite{kundur1994} but cannot optimize multi-step costs. MPC effective but requires accurate models~\cite{venkat2008}. Single-agent RL applied to dispatch~\cite{zhang2020, cao2020} but doesn't scale. MARL foundations include independent learners~\cite{tan1993}, CTDE methods (MADDPG~\cite{lowe2017}, QMIX), and communication protocols~\cite{jiang2018}.
+
+\textbf{Gap}: No systematic evaluation of modern MARL algorithms on realistic frequency regulation with safety constraints and renewable integration. We compare CTDE vs. communication vs. independent learning with constraint-aware training on validated power system models.
+
+\section{Problem Definition}
+
+\textbf{Agent/Environment}: $N = 20$ agents (5 batteries, 8 gas plants, 7 demand response) in IEEE 68-bus transmission system with stochastic renewables.
+
+\textbf{Formal MA-MDP}: $\mathcal{M} = (\mathcal{S}, \{\mathcal{A}^i\}, P, R, \gamma, N)$
+
+\textbf{State space} $\mathcal{S} \subseteq \mathbb{R}^{140}$: Bus frequencies $f_k \in [59.5, 60.5]$ Hz, generator outputs $P_j^g$, renewable generation, load $\in [2000, 5000]$ MW, time features.
+
+\textbf{Local observations} $O^i \subseteq \mathbb{R}^{15}$: Local frequency, own output/capacity, system frequency deviation $\Delta f_{\text{sys}} = \frac{1}{68}\sum_k (f_k - 60)$, renewable forecasts.
+
+\textbf{Actions} $\mathcal{A}^i$: Power change $\Delta P^i \in [-\Delta P^i_{\max}, \Delta P^i_{\max}]$ MW/min with constraints:
+\begin{itemize}
+\item \textbf{Capacity}: $P^i + \Delta P^i \in [P^i_{\min}, P^i_{\max}]$
+\item \textbf{Ramp rates}: $|\Delta P^i| \leq R^i_{\max}$ (batteries: 50, gas: 10, DR: 5 MW/min)
+\end{itemize}
+
+\textbf{Dynamics}: Swing equation $2H\frac{df_k}{dt} = P_{\text{gen},k} - P_{\text{load},k} - \sum_l \frac{D_{kl}(f_k-f_l)}{X_l}$ plus stochastic load/renewables and N-1 contingencies (probability 0.001/step). Dynamics \textbf{unknown} to agents.
+
+\textbf{Shared reward}:
+\begin{equation}
+R(s, a) = -1000\sum_k (f_k - 60)^2 - \sum_i C_i|\Delta P^i| - 0.1\sum_i W_i(|\Delta P^i|) - 10^4 \cdot \mathbf{1}[\text{violations}]
+\end{equation}
+
+\textbf{Objective}: Maximize $J = \mathbb{E}\left[\sum_t \gamma^t R_t\right]$ subject to safety: $\Pr[|f_k - 60| > 0.5] < 0.01$.
+
+\textbf{Assumptions}: Cooperative agents, partial observability, 2-sec communication delays, unknown grid model, historical data for validation.
+
+\textbf{Data/Infrastructure}: 6 months ERCOT SCADA data, Pandapower simulator with IEEE 68-bus, PyMARL2 framework, 4x A100 GPUs.
+
+\section{Proposed Method and Goals}
+
+\textbf{Candidate Methods}:
+\begin{enumerate}
+\item \textbf{MADDPG}~\cite{lowe2017}: Centralized critic $Q(s, a^1, \ldots, a^N)$, decentralized actors $\pi^i(o^i)$ during execution. Addresses non-stationarity via CTDE.
+
+\item \textbf{QMIX}: Value factorization $Q_{\text{tot}} = g(Q^1, \ldots, Q^N)$ with monotonic mixing. Adapted to continuous actions via NAF.
+
+\item \textbf{TarMAC}~\cite{jiang2018}: Learned communication with attention mechanism. Agents exchange messages $m^i = \text{signature}(o^i, h^i)$ for coordination.
+
+\item \textbf{IDDPG}: Independent learners baseline to quantify coordination benefits.
+\end{enumerate}
+
+All methods incorporate \textbf{safety layers}: action projection onto constraint sets, safety critic predicting violation probability, and Lagrangian relaxation for soft constraints.
+
+\textbf{Method Justification}: MADDPG proven for continuous multi-agent control; QMIX tests value vs. policy methods; TarMAC evaluates communication vs. CTDE; IDDPG provides coordination baseline.
+
+\textbf{Goals and Success Criteria}:
+\begin{itemize}
+\item \textbf{Primary}: Frequency stability $|f - 60| < 0.2$ Hz for 99\% of time (vs. 95\% baseline), $\geq 25\%$ cost reduction, zero critical violations
+\item \textbf{Coordination}: MADDPG/QMIX outperform IDDPG by $\geq 15\%$ reward
+\item \textbf{Metrics}: Area Control Error, regulation cost $\sum_t \sum_i C_i|\Delta P^i_t|$, constraint violations, sample efficiency
+\end{itemize}
+
+\textbf{Evaluation Plan}:
+\begin{itemize}
+\item \textbf{Baselines}: PI-AGC (industry standard), centralized MPC (oracle), behavioral cloning on ERCOT data
+\item \textbf{Training}: 5M steps, 32 parallel envs, curriculum learning (normal $\rightarrow$ N-1 outages $\rightarrow$ extreme scenarios)
+\item \textbf{Scenarios}: Normal operation (100 episodes), N-1 contingencies (50), renewable ramps (30), distribution shift (50)
+\item \textbf{Ablations}: Coordination mechanisms, observation spaces, safety layers, reward weights
+\end{itemize}
+
+\textbf{Feasibility}: Pandapower validated, PyMARL2 tested, IEEE cases available, compute accessible. \textbf{Risks}: Training instability (mitigation: gradient clipping, target networks), insufficient exploration (importance sampling, safe exploration), scalability (GNNs if needed).
+
+\textbf{Timeline}: Weeks 1--2: Environment setup; 3--4: IDDPG baseline; 5--7: MADDPG/QMIX; 8: TarMAC; 9--10: Evaluation; 11: Analysis; 12: Report.
+
+\textbf{Expected Impact}: First systematic MARL comparison for power grid frequency regulation. Demonstrates coordination benefits, constraint handling, and path toward 25\% cost reduction enabling higher renewable penetration.
+
+\begin{thebibliography}{9}
+
+\bibitem{cao2020}
+D. Cao et al.
+Reinforcement learning and its applications in modern power and energy systems: A review.
+\textit{Journal of Modern Power Systems and Clean Energy}, 2020.
+
+\bibitem{jiang2018}
+J. Jiang and Z. Lu.
+Learning attentional communication for multi-agent cooperation.
+In \textit{Advances in Neural Information Processing Systems (NIPS)}, 2018.
+
+\bibitem{kundur1994}
+P. Kundur.
+\textit{Power System Stability and Control}.
+McGraw-Hill, 1994.
+
+\bibitem{lowe2017}
+R. Lowe et al.
+Multi-agent actor-critic for mixed cooperative-competitive environments.
+In \textit{Advances in Neural Information Processing Systems (NIPS)}, 2017.
+
+\bibitem{nerc2023}
+NERC.
+Frequency response initiative report.
+Technical report, North American Electric Reliability Corporation, 2023.
+
+\bibitem{tan1993}
+M. Tan.
+Multi-agent reinforcement learning: Independent vs. cooperative agents.
+In \textit{Intl. Conf. on Machine Learning (ICML)}, 1993.
+
+\bibitem{venkat2008}
+A. N. Venkat et al.
+Distributed mpc strategies for automatic generation control.
+\textit{IEEE Transactions on Control Systems Technology}, 2008.
+
+\bibitem{venkat2022}
+D. Venkat et al.
+Economic and reliability impacts of rl-based frequency regulation.
+\textit{IEEE Transactions on Power Systems}, 2022. Hypothetical reference for illustration.
+
+\bibitem{zhang2020}
+Y. Zhang et al.
+Deep reinforcement learning based volt-var optimization in smart distribution systems.
+\textit{IEEE Transactions on Smart Grid}, 2020.
+
+\end{thebibliography}
+
+\end{document}

From 7fec25cf4e49fc34c1de7b4325f03f2dd7bf7406 Mon Sep 17 00:00:00 2001
From: Derek Smith <60520818+Br3ady@users.noreply.github.com>
Date: Sat, 1 Nov 2025 15:07:56 -0400
Subject: [PATCH 14/17] Rename proposal_3page.tex to Proposal.tex

---
 proposal_3page.tex => Proposal.tex | 0
 1 file changed, 0 insertions(+), 0 deletions(-)
 rename proposal_3page.tex => Proposal.tex (100%)

diff --git a/proposal_3page.tex b/Proposal.tex
similarity index 100%
rename from proposal_3page.tex
rename to Proposal.tex

From f184586890b7531fdad62fcb4f4dc15c2629149b Mon Sep 17 00:00:00 2001
From: Derek Smith <60520818+Br3ady@users.noreply.github.com>
Date: Sat, 1 Nov 2025 15:09:24 -0400
Subject: [PATCH 15/17] Add files via upload

---
 Proposal.pdf | Bin 0 -> 249416 bytes
 1 file changed, 0 insertions(+), 0 deletions(-)
 create mode 100644 Proposal.pdf

diff --git a/Proposal.pdf b/Proposal.pdf
new file mode 100644
index 0000000000000000000000000000000000000000..6a8857d60df98a7b7c53525f043a489ffb7f27e8
GIT binary patch
literal 249416
zcma&NQ>-vd(4cv2+qP}nwr$(CZQHhe-ecRg?frJMe=^C;#q?EG(tXnxm8a|JB2^F(
zqh+LHg(96_9$klGBVZt~H?o4_;en!;F|{*yu^?dLU}7Tp{}vRzn5B)2sS^Rcn2n)}
zsfekuy@@FlA0L#ni<7CLEtJP*t>%{9wiuG{n)<yVurmVul5CSFWCAFTq$7?-AQ7w^
zdcu!DB5kBK9OtQM)Z>eHRrSO{48|7x7ZXW$wY_S2mD~%&*)HPj7V<;t*<5xqz8|EJ
z_*fpj>bcEr4#{&$eGWxp9#UX4We)b6QN^dz&#B%THNVr>3Rgm)cj?q;ouzXxuZ3TI
za~s{2_ixYuMXu_VZ0E}N#}t*^wwgblOD=yspPip`3BI=>5{EVaeF%)_0kto`il#mr
zD?IhhQxclZ`B|i2i|(<lHX80sm@uWd6us(Xx$wk4Xe4~+@;oV&Io~qvAIKwyUoY)f
z_-CwcO<jjA(bYHj&q)5?R;d#14BxuK0s@%aR+wIj7Ux|3<{XuY_-a=Sinq-i+fd9e
zrfQE{=G=7L{85_w$dwRxXSyVhx@Ox<=c7){UyU}|-d1JT8mb%`$=EU#BUaa1_uk#4
z`5UWP{bD4U!}5AhEB*H<Rb<k*5@r!GR_ja<JuHz~E{R@+5uRL~+ENrCu!`p+s-$^o
zxw2@3v#`AN>0nz9+pxAJwVS|XFUw<bmev6@U-}fpxlT_C?rYkV+qzL4DyP5dr#$VZ
zA`90wOKOD_-MzGS;d*GR`yI2jbnQjrms5qp{0`r(bugv;O!W8=heQNA4aj9bc%A#L
zoe<8a4Y~KAv~LLny>^xytybLR4Ovq)-;3)ALj+F&x;QRpl(C8R*0jFzloygC;XwNw
zQpkY7S7cwd=O+o<gNUS57i+r?65B%l@F0(M@WPSN?3d9E(%s{rQ{N!4YwPxxqoLWg
zK8YR)vhXi#!meF2Gvlrsv@&s<0HHXuam6>JM)IP@Kld&C*-ywpDoj#8qJLxo=ZlIC
zNGuu?nHSa9ZA2hZ>$sOq<SP;XDo)Yrs<)B`NOXhxK|q?~Obr2c%zT&J%EsNHdriy7
z&(p@QW1v}G%VYg!I(iHpT17R18SizBO)&+ck9nPGvCtub+_O`!+ZoKve0U&sBplA_
z{U4m+B{f#H8&GmsSpm)M$l*|=#~_vEOCEMPt{7lsO^#2@@dT=eyDwByJ?AeZPaCH@
z+Px1EM1%6dk0@~c`4aOutPZK|!*olbsWk^{x6G=YxxR$e(mNvq&Z(~3(g}nWX-CHZ
zW#HWoZqTVC6`aKn$<MNr=g3324kX6mk>x=mk*nsB9WRf!lC$H7DCa-2gCQ43Jei8L
zX%NSt>{Fe6AHmCef<rB--ZO!DJsX=ITF%oIqL))X<&H@OIf&dZ2=r(qsoI3nX;P$G
zN*UQ;2}w#)6plhGQ;Yr$tfGXE#eWZEPWU1S(bQP&idFk>MX4sEN=uL70sQ@GG4W|F
zf_$W(#sJHc?U~6PjLu&d@}8+ZF0CgY{>%Fv`aMveX!+AmXCW+%bVvz6J6<#E)LRZh
zxd=qFYaaE9M7bFqK;J(nKgyE&$M}!9oZ>`K^&YW6U+eMV8Gy`Km)<!4sD2uOO=L=a
z)v~#Ve!`x!lSb53DEo6Vu#p!LCli<V3wWGC;S%H@om=?HkRfr1tbs9EsxN|Iz`P;*
z7CZ0I*wjfN0;(_Z{7R$sjC+6w)(``8XX$x&TIJ)R1n!u=^c_VqwpV!qVFFs1+e96t
zXn`h^p11*4xib<n(ph6=`;k1<66r=N9}PKR`?*)|QQIV0i9B`b_O;+}x@X=XToKI(
z^U8iB!bymvYyl_NTvwix0S2iu$pE3K1;*B(r6C6Eajz^u3VZNBCR2qSFqwvvHAu%D
z>aB!qLcBD^Az$?~*)6O^@!WrHX^ktm<#;Ej8}k<rD<nre)HX;;^N31IDWNH<i_5*V
zGnWOiZTM6L%8|oI>#`m)W#~;;VHLCsatQww%BgSpzYm37tRg|(j1ef^-u9N1C0*5v
zPd(z%Km~FcjS7hUT-IRR-4}uk{su1(kM^m@$fkEHQZ%6wWH6H$Ogx{hi`NB>@3xxB
zJhE1uG@;N@5xyvUSF?@tMMmxcS8l8rFsZkR5*IIO3s+rJv(Pvp1Y7D+KJRXb7ajl+
zmHG@sf*WD)sZ^ljj?^~S5CJPQ-+;2h=gY`A8S3{K8l=l6ZeCnQ#u0p1K{Y&^OtcPe
z3exy#S#zp%C+h$aT(D*fbXpGjAWaTL)dNZb%@P9p7X1OPDBr11^6TaiM12x!WoJ&W
z-V0hVW2egs!7fOYglR1%n;7NbpXBhQHf`M$x67KV$*)SX79cUp<*mU_X9tWYW!rE9
z&vP&vfoOFzzcvpaJ<MNgNL_I7>gVqO60etC?JDxZZ-7o7*s-3e&|}f1bRCITWn1Lf
z6E?j%2&)K9FfLH^v_CyS7+9*3bVW}ZTH8)5b9mj=XK8Axpg$TQaDYeDVkSIANpNvn
zhrj7NVxc>2Yd$NI23keFbM>340e?SO@xb@Ev*s0u(=(~y&3*Bk2IhM*$U5B#JRF+Z
zo*TA>JiQq8Lrv4My8P>m?gxgfqJP$6z=*K=c%>{+J!DB?e3u6Ate`LR{1us%*2niD
z+#wl7m@%I*D;4-#OIN3Mwg;d#3u_A`?3cy3a~-2he+h?Ycb)9joLWj4#dW@!eK@m;
z|0sq*TxL?TJ6Hix=eu)Ngnl_3O8+?;3*(jjP1_{HgFxjEA*lMSCHU@B?!utBQz@e*
z<|Hg=U_!N1Su?j*`rq4D@$($Kg;r*<_|H+(6OLbxyOwSEarE<Za2*Fe9pt!7zJ2dv
z>@K?*Q?kW#cX`-W>L=d(6Sv5}?gwwDATe@wv_}`3t+YvuSZBuQm0KxT;OzedGCbq<
z>zCy7vj+9vNISJ-keQe(?#;*lw7rJp&qQzbyUbfuqlyaB(TE!G71s8XzA4{kU?@f?
z^!|v`9+Qz#2d|-J*#*JZymULS#TZz7-nOvIo$P1v$4zFi+A@Y8>*hHTvyHs2usd>Q
ztkO;x0LIM_9oC33iB}Q~dCxl&wl)y<)y}4JG3~ZDqVUYE80gIdKV3J$dq&1D4BW6&
zPX&>(VlhF8g`n%MG2c5*%YJ5!{ujSJ&@Ci9NPSi3XZj}DDY`XEziUv)sbo9hw+7?N
znAjo#$l}S0a@xc&B3}5WPJ^X?v|8C-G-b>N2evDMLIPy6z62Y~$!%U)dy8MAR70X4
zk(;aAXBwHEAP#nmkAC-STnXyB?gQ+!VY89v!u{Z$2>NVK!Yylh;CJNDJhErjJxVnv
z9UT0HIg+(!iJ96F_QzXK(*Nw0&DblHZnmioYs@8-4SDAwhhb@b8aKvKMqPc2eIFTa
zJI;GFuST4w`51QkF<No4qu-<4eiC{FN2$7p$&Hr)b3!`7IN>3<k<tPp-%BA|aY1pD
zWIusH7zD&BJ>b$^(~4h?EfgAmc-Yw7L@#SiQtlU2AX^H<kE|+LlJ(nN06CM1{xtI~
zab?qGH#^3({VdnGsPlegd83Zz)5-(g=>~?JRfsaxmB-z+a27q{3-eBN(ewr#NUED|
z{2^p>88f7$k?tMV@~@((IQkNQcE<S=dq_u*9%V_WKnkwkg*?ZrQKrv>5@(t|#V@Ee
z>Ditx)L33@iFdT`vZRAlhDG6<OBYl*1~^-$V`Cd^X{v7WT7eqD`OAY+<__GHs@&JA
z>`TY1kDH&r{E^MI*u<CHrEOQp6(Q3)|D)F-du6BX`Z`~rIC0{9bzXic_0lL))U!ME
zE|ZXX%}_<g|B91O3Y&E1PggQ2#{Xhkbe5g`4><2fViL;K&gB0G;{T%m7p9pQ|97fp
zVPRzczoqJ4ZEfc*u_nK_`h7kdR1NE)3~GkiiAp7sjVqPoe{HjpRXF?tBqW1`ke~p*
z6ZLzynSczyup*P2$B`W{XNzz3-HFq6<`dT^$s@nN4iCp0(3NIXYDtqDqboGYjfXyI
zN_fs`vvg|BZwt|#Ns6lWk1sxhOlY~dw6k)vXZUn}KCeNHZ*x>>I)l@n_H#E8BbbCo
zB+}i0=4JpVvj9U!DT#I?Z%?4I&FAeRB!LqKG*{~0=H$}7D~mPxSKR|J4icO8XP5uv
z2oK_^Of&U@$Dqy!j3i4i)xb^-8H3lL&6iybxOvntaod^+m||xiLUv}hK8TptkY;5@
zd<ae6Ze0U3bjC1ex-bK!MqnN}x(XoUAz`!!G(!y1bZmFg24+d3v$lLRLEuQ6G?bX;
z-3LK}MJZT~m!>hLlQ|M=u)kJsQ#nIqjEh2~#fOpuAO`U`3-KuP=g{vdJ`L9ZYE95|
zS4Q!>jk;-~$pns7f)rqQU>j1#HM4+Njp{luU%dC9EolN{W(lwi0-`(@rT4uHfxZHP
zOKnP!c5&|YUU^WEjhNGn^kD>|zA#OYGJxt|?TW~=-VuysfMPD@KJWPRftZ<)bkVnp
zX1A)?dMe#+Zt7qkL`w+>d!O^c-;;vcx_}-L#|iQ^#>}A`a4H{Nrnd4TK+n-2;#+|1
z-qeSG`Q|LqPSUT2t{3b&Yue;zpk#td%%{~8TvO|Tj07QRmUE4SNW9E?%O#PFz)Krd
zo(<BVGDt^|pjZfbwdec-iJ8ZW+O`YKnDKN`J-e%;$>1_UD~*6qUf_O|LH$b>kZ5}o
zBvo|TM1L&lJWDQ6kE^4SOVD%kgMDU{I_H9bzys{+Hsv+2h6E7KhiPIVKB_X}(LlJ9
zp{gta9*Zmonp@b}_0RS{14e?PDzbtXetTGfW2X7FPYuMU&`fPHsESXmm(a2wR6gBQ
zeW+M<fq@zbfIf(Qf4|d6NCtvw%ph?yc|^s?=$~ZpY^aZYG1mbPwR-P>LBbT33guOn
zZ>8-oCoV>l`SS|y$(b{NVZu;c$OZ|@1rAS6C%)h(Y(VM#b4U$}n>WmPU4UVfaoYRy
zdc7aVMo50WepDp4Hvlp-V-uI#6)TnAi<4|N=Lft4Z?R;ta#mddiAhx;H&Ie6cZb84
zGXY&c_~MNx=x?7mLLxGm&)WyR?kSDwk_T_TUa{Opy?R^92DOHlwS;|KR98ls>^B<{
z10v)CN*_b+E#k<A8o~V}Ku7Su?-b|pC=8#K-(Le_iK><>$}xrjW&;ZwexO=qvR{hW
z#x)L)vVFXDK`6k==&*-oy;CA^kMl=bFo)mMytj7o*+P0^uHwi?0NR*GHp4N(z;h;)
z`PYE8Yd4Seg$m;$c+q>-8h=i@&5*lK->$iiEq&J?r<!FuS&iO0ZR}&}b%VHCT;b>i
zsJ6twHF;-&oMn0wavbN$UHa}5+*U)AOrGiOkw2@pMjDT7^TLC~0Flf0ghXz|RbxC3
zzv&&~Qsf-tOc+y*=Bf;a;OB`RWVMdNHS5rL^*$}le2JX+9o(LypwL9O_l0{1YDy-K
zSnTUz<fZV#(2#!b!qMOp``SNFSG`aI5@jJSpb6@(H=Z2vPv7aYgSgMP-$?NZJB|(U
zh|<LeOZ)N0D!<;rEHDVoBJBH;XN?XR;fQ0VdTgV+M4GJP$&<>*Ms&`0h?6wJ{T^RC
zJ;|qg8)cO<0fbP`%b#avC)RBHDAQ#Pym5@a2GfV7fDl%?=T2U++7jm>iZ&|WSwp2G
z8pQ-Gkyd}Xamo}My7n16N?)wxm=?q}hGYh9%D}+_`$&jfiD>vqN@GS)e#sysl@q?a
zg_JM^#SkmIStl_9VqD0O5zUwDI;@62Xrvg`PG#N^Tglx)fOMFWoauwcYr>w1JOUaC
zix3OG5@;btGaW`tu#f)G#452K%Hg+`1G0W}1G|;KwcSO5x}}WaDDDLbf3=Ez@1sMA
z$x2@zBo%T%4u~CrwC&0;D7mBb*qp<-C<b?sB&GpYg_&Rgz0nd3F-m!821D-CrG=&>
zY2D$q=v=t4%%fuD&;`63$~6LnxdmDQ2r^*qQk)u5Qs8}J%XRw0I!83lLHyD{7-%zW
z7>FmKrf;y6BP|B9R)c(FitPYEQ>A?b4T?YkQv{2AUSI@~cV8__#>!#`vYgn2&Oh8t
zX&$aXFA^eit)dZ1L60n*(?4mUrfhX}Pt#Ajr130*;pj*R=d#*5M5D&XZok?#nQv7j
zaBDQ<?%Mk8sX?sr+X;u^{K~N-Nd|I{EfgHW4M`+ai-HaKqkWRsMkiH3tpw5<^z`b1
zpv!3FAn?rzbbIv}Fr5h**;X}a=YYRD#Nja8yb(P82gb$>J=b}dCwgJJ>NtGxsn`Jo
z&^G`pYVd`J$YjnCNXP1vz7}uDVN-xE!Ul87mn5M|mv$^(ncX;KUyn!MOel|yn^2Ba
zkjyAM+G6c07>0nkc&}oz9N^3rctcPTn4n}eq*CxlVsT)NFm30r72|^e2?6B~m29~{
zT)}=?0UL!Yy@8C(m&YOLwD_&oq+d~6U(?Ve0xSn~<=_pozRAjdkTPZj#sQXka`y~3
zMQ|6V{&>S0ulolzE|}f)M<*ol;|SmRw9hIzB!Mn%1&W(-`3iS@YIfY@0}#j%fFaAM
z7YRF``B0!c$p2K-tm|tIPAVZV%6=URa}Z(>a=PLo<PPZI7~w^rF_=bTwmP29rapuo
z&y5>xrsU5#qZ^9l!19e3_B4M>w^RJl04b67bHb5-F-)*fg*|x{^dDG$SyE(IJ};Fs
zw`aP2ghO`2<Ukq?PRS3ty=K=15H5k9*v^4P56yvwF!Uu3G{e~}a2fVp!ER)6*t(^4
z7sS)HlyYNxv%ahDojr+ZWW&Y6pSV&O!Ney=*TH1ThJRm)L^1ARiC=Hu{rfjK6NjVk
z@Q3h5HIMy&!)!?ko#N8{pux^%C~%FxTC>VFZ0&ZmV##xHdMcOD8xo?#$5ZVjEA3dB
z7E>SPyB2}Q#7)r145Gd$)fiP>ve#@Yqf75s(PBa)bOem63OO?9vWzUl=&EPH^zYhE
z7EKbLk5n6HU;4W~V3P*D;&B~y)2V|81Q;VQ1e0O~T^M3z6L(Sq9xs*FoFDvB+dFCF
z)!RHku+tR#=+wRN2`_=<2HGg>OOZ3qH>z(`VLKXkX&V>uuCoCi3g{A*6Gf2b$PAmq
z(%54=`F*1yuHVm~oxg71&q4O^Gqtxn>!_}%G{Q8P*>x^24J+%y57x_AMs%}xVSG%i
zP(5BQpP!d?CTtu#t!?l@EZEi$5qanO^Kg5AA~%f;bM0jS8>{@=pHK9s?+|mAb$WUJ
z`tgL#u#6lNbPGg*l%Onf66i1jNRvTmu%NJ|i>ek3kQz2M7j{)XGkpylT2mILsPx0{
z>8E8Y<zP2H|G9;J)7VU9<;IIuClDDR3tw7W<b&taGGUr`K}2aPXU~y7P-xypL%R|u
zHWw!+KcyFzrLvR>(+_+93UYfO07Q>m`PeBV?%$}l&8nnlltQo=NifX(WZ4I;qhK@Q
z1?FVtFB%*4<El#qrNw{!uFk%1vvl&h-!9fY(ZX|(LoJ9LqQ>fj!&8Jjr)5*bd6zFc
z(F_>T)S6ZgqG)&p<7x*1fnw`}&@d|`CmJBp89M#<GPLtNOyCQqx*n?hEISbr0Xo3B
zw64~zUHX+sI+SkQ#*~IUtQz1c9MoN}!H_>9!EA-8xh%5mN?B{$q{~vv-s}-WPt3|e
zUq4`CM;}KgFMp7Lu9c&u5a6-Wb~^`dxu4mn))y`U5F2)#g@zM5`#Zw~4{xBJzDY$3
zdX#TpKZpd1@GKLnIy0S&z3j|<Y_l2;<+(0~VcC!Ig2Gkpha)0PstB+6%r+oaUF&e|
z50idTK=ntSj3-Cv!znogm$U~f6!ho1`V2_jcdTyPyyT(ZP16^s5sXuQHlT>wSJGbg
zl}m-2b6j5Rx=_kK(QR$UG>{s!kHH3a-oiSPeztBo6sHdos`V)}`wHzczX5GCJA28z
zo#2E6mY~cH6LDcQASup|>e<a&u7wIUdS7Q_A3#+jp@r;9&n~j=q|AmzP(b3)&lh4d
zj-qF4jI-r@)t3cEH2Ig;#Nw!p#s-KuahMYB)hLdXL^5f8{lzDMzF!^CVS=*Adh)`C
z55&1=N40bJ3Xt2iCo*+GlV)60TCU7)NDvF*39ou9$7%U8p-_26d|Vn?Rqdr}7f!Gd
zn;4K4^d2;gZ;-1m8zMpDjC?saW2a}U9vVvSU<$G%B5qB)+Pcy=)kC>@(&jM87XQ0`
zqLx<QCzmt(yF;fK;C|Ld0~C~e0kIkl7QYHXrvmj@dnEH59_y)Om#%7C?1DSjDhi>G
zfTX;VzjO>NP}-=B(t=8WNbC@-S;Q4_Tya&SIIe6(o47G%t6O-O?V|iPpVd0K{edfU
zs;P{TBmRjK>bPp)YC~XUPOo-?WUY-RX`(r;h@*3I`D7c`bySrO7}<r?+S+(4|KSST
zRy^IV`*q0mcPO(EZLX}_YS%hrJ<5eL$B6+bSJQMpuDg`>@2m9aU3^~-CX{7lE+dzD
zfe!=2S8RKjJa)S6G;^0_O^Mwjb=m{fW0n|td(&yb*!B<Jwe=V706RDXEFRi1)^86@
zb}K_V+CmlpxIOnCX<nX>uOoHVwX@}OU+TgAbOH2HoEyu-%O>xl%Ay_V70`7vWS8Wn
zCvqoPPT$DkE!VJN;ZqKWrE+qGL0*IU32bd8%9p|#)?!&=9s)hY{myNQ)Pwlx86%{#
zVv1*7V#*a?3IS)=vUV}}WOt0ueq7J)tcQ->^M^IZL9R4iVxPu~ajkT%uofqJGgqsy
z4h+ejDs9-ajLCk3LY-rBhjVbn<w62Bm`pj~c|R7!*`7u1@}aZcWx0Yys<$-r=^%Q6
zr+8Grw)EFBPeuJS@Z%dO-yo)-luwyKZrw3gS?KSJC1qlU!TEOIg(}Ey^WwL3XM3v3
zvNm1HA}3MN8A#(|XVK@qNcCc_sBXJPD4pnunmBpLZoR!e+CghMKDN~rg=c>qL?_A-
zP$YZZJm7pD>D!7eHs5ePg_ObEm^gHBKnIH*nV{p5V4o0#T*BcfQHI+WHqO`8f6AWf
zj*J>#gzjG6wL>00h{Pu2CDKI&oo&vO=N-2BYN@O4@;5o)YG*&qlV0ekjru=)HH@#g
zOz2j{bzd%QPBp}p@5NeBDclv_e_Olw*i)Rv1-~h*-`U)Euddu;{dCl?e6l=K{2HK$
z-E6Gmzp+xx-nu<Pi18_ftCYIY7xERXjLO7^&CjAATkXb^ZaVkL4PWL4zE8Rz_Nm63
z!aLCNZ%AWbzocq`;Vi`2|9eu0#=Yy|8S@F42kRYq4-`SbTGgLXzHi@B)vJ#Y1mBsG
zLE)MuJ6vye=1pBz2E@~-gFdrIbJ#v6LkfDcX7=i=Yb(1#<LBGT3fG|d=h8T!;o}U%
ze;3vo*S{EY^$jeQ0{b(r0;-YJ6&6W}6v_M@2GAg?y<Ie~$iZ2Bx?u}vu7_^zY)<iB
ziQTfjVgTTNwKLYUY9OBsdY4e7{gmppAB;7-y?=p!Kp%e*aOe7kCkJh3DxvX?dyC;c
zKU_}J*Ww!37|8eCG;nRc^8AycOk+gJ25%m3u-I)^ix6J&OEBz;tNGmZwVp4j3@#&E
z6R%h2jw%*ryC5zyYP;c5+CDbP0@vi5Fciow2V9aVLh^$z;!?h|$OqQEoFV-UEuF)v
zMrvW~+1X5K;Y}8A0PJUBH%XmM%V0fRc*|fYhKwoQ!iLh1JC$?ZfbZ<}aLiBp=al)s
zH)ELp56u`R2G0L$#W|xP?Yzwa*LSA=Y+ALfN-SC;!n;Zyzx3*|dh^Ip=PSbmu2s~n
zxlkOQl;h$50|2rj-^n)K*+Gfm1Au=I=)$2rH?%R+eSiFxKJI>3@60z9%yTet{ot8K
zDu@@YIeA17x42=<`E~7kK$G+L@b1`t{o|U1U(-mO(J+}VpL*@O<<fEZ<BEbyF{X*o
zWV8G|z8!zG-feZpH`lY8($z?P=7%DxR~yAC-y3V=qVCS0+~9ke;+YA3fNSBVty!U1
zoIF=G9$NpbJHcu58|9yZnHuvMMzxso{4*nW^#WGf+x?Lk*(VXqW6Yc%9=MN)>A*XO
zEI%Q~e<Ahu6fw+sCI*cb@;z@x*5l2iz*f4Qbn`Bm&Ojhx9B)(OLg}sg8zU9z{@2yz
z({WNVUv0<VtV4rFLT6IWg{3uYiLX0RF2#9LRQsP_%uxcY0|ur-mC?X0<q5coOZT7O
zS)<hOL^2`T&4$2Y=9d#05S%gRK^XG9cFYTWE46cJ29uANUy@yue#8$98=^0Su->_9
z6~**PE9q=!Yg(=ntoBkd)>FK=L30BdKBRf}vRrjO4K$5m5vm{5XsDAV9OxUe32$!T
z9~Nz(&f)igEG`lFICw2A@3XI7?GPA#y4UJ*OUTq?3vFr(8y9nH+EsI@Wh8P$!M8iN
zGdP3RjK%vm)-YOZcYDjFRgo{bNugSWv?f<LY9z7qK622d=77$c8l#x-R6G!3@yiZW
zvaHU-0m*yn%--0L=eba3YxS%Yr57b~>sp21hRz_@iT=^3D&Kn&4*$kpat}fnsoubq
zNdCQS+u0>j`qp2Y_Sac(#KdkqERwZCn&;^%2uRvQAf06V<dh#8Cx%963U;*z?Sx@$
z_u7=WVqM-1?L=`T_gmoR@LM5`zGnz|+(ktIXo#-STN4=c35gG!{dk>l4|2BK%BCw1
zccn)dZQbE0xR<fdGDMRq=AL%E|JBLL!eI;pqr)D4VDECFhVgO+pfKq2AVtJn22KUX
z(Vzp1uSOg21qg+&vcty5nHvhzl#vhO0WTQCf^6=w!2Bupkepg2Cg?QEi&SsGNx|Ns
zakmJ=5_lk;PFDe>PqH`ZCtTdOL)utslj8WW<l%S5##Dt9^#iK{Tq4dBTn^9;+=*94
zO>PaNv*?AynKT6Kn{#+Vn~DJmH{?4GF~LArgkgmz0Mjk1m>}ADB=J&XAa1ad9Oa3E
z-Vj0D+6DBL`Quo^`*j_S5C9<HG1No7u?Vm}{X0oDfPmtVWk<jSrFxeEeMU_Rt5~xl
z7Jn%4mYZx7SOBG7W?5CZmo-n5>KHTu<k~^$F;Sy$DRZ!GXcwC4E-6&3*YbJllaB8d
z7)dZY>S*I@0}k97wG22;bQig--RfMR--R-Hw`K1EpYtgqp0uq81+E@YG9aC_2y`2<
zcUAd0O^haD$U;_WY>FpT$`3Yr=7?0{>9!5xbDF&F3o{RSUtJ0-h@oUMyzkEL&OJa7
zNqpxPp1yMB9Zt3%Tqi(yFBl8SzyimCq^&g|7z-U%wbrL|SDm3N0fCedH5t#F?mcA&
zgi7%aIOQh#x7hUy8ifmIv&u}9Jp!+o2R@J?j<8!f=;bZq82Jr^UOS!#i`asNrq*t6
ze#0M(0)4Ra_=}9zOL;#8vPi)P9B*pA?}&D6=m(_70=dh(nji4`c{ZOCoC&o@5`4dd
z+Z#PGjkzD=xHEHQ8PX;z&sRY<Z6pLT(N{j^5y+s4$}Xlv21Ng{zp8J$rldM+mv;8h
zx#Kt>;$-B`WBDd5@a^ucqGxlpxKwY7N0U>%q|w|gx`)-zkOhS$QDI}qUEg1wWD<^;
zsU*^8U_p%^jR)z?3uVdKiu3gf9Cy7fy9;yZ0zd1CKn!M#Ae;rl^T#db_-U+2L$=VP
z&9nXs`JEC8qDQ^%y_E;>704lKb&J!_o-2e1o*!@qi3rU%Ic=B>;tgpve;=6UiA_>+
zDT7$;nc?Biusv^>&I(q|RIr<1q9_P&%An%#0PGSO?#m#=tV&p?#hhnUxQ^`Ra#`#6
zk=UFhX<o~ag%p=2h(j!B#={x!q{4V6iau(*5PF#&1~b&8V*Fp7N(Y~u_P*Hmv?l|;
zxQV;f2~HDS$&Y(i2#9xU>9z67uCfeLd9n!ya=^3Xh8arM^CFFg?L)B{*qsG+QDFPk
zHN_hC@NDh!C62)yc2my@VN6)$WKeKkS`x~V19HCLxS0L<oeI-?a)l$L{gQM;+69NT
zE%U?rOmyC^o@qp4ZLqHGB2Ei}4eP_*a(ve@c5eQX2WCPKG~BeRA`r2}03g00XsCC3
z-a5iDU^rLWaA|VzMdyoi3xM#kJT6b{t49P6EhP?W%%<9IS!c0b>D$<6*oQ%UVs?XU
zcGg%Vh*?C|X=2Y-?s(zE6M#p9Ke!jb(g--)-0VUfRRJ;nSK>Rl<ZmNTabsLn41Wm_
zYEDJtexK5UbC|||w-Mu}4WG(mu^c}*aVT*8_i;QRS&Oe6U*QK27npPm*f7aRakIsH
z-g%{b9>?UitL3D=;P30IvT9#&tcf?u!F)s<gmxW8q4GI6c5872=*H<{dtsS5M_%|<
z9}4$5A6qhQ2G=-mssM*slF+N*BIAC$gOAt4=)lYOD4o@iT3qXuk9Qnaa?`ff`ny->
zv6u{O-lc`rZeGdl<p{X?2^l_IfgcwijqfXgW4gyVfYyOHC4Yp<bO*LKrwoe;JS6k`
zM@;DGIb96H)e*B?Oq@G<)ZGtok`y8Hzq1D`^ZzD-7zr4e*%|-mvlSBoD<=!*|78E|
z{-2P5jf3O=+84#TfV1T2uCXnO-3cPF3%L}(?OMAO!GjHF7@D<IL_#cBCy2lrBJ_$x
zpz(GQc%jou5PJ1V=JDKp@!kE^zxZ0OcAe$)*7v?`nCq_{7_c&&syhjB39KYS!1N~o
zXkaNUD7PU10UI<#0ASF_j0_qA4G$FhZ$)xU7f`UkL?-;ghd@C=4er}iA;7>@go^-d
z?m7rafFK~Hqa`JP1p^u~Xk6M8Ad$jW2>BSu1>^w~#EJ+kh~Q93$or#AkD>-wKfNLC
zI~{|Cl$4OX`-KBb;VGa@BL)FH6JV%UfY%ifgTP@B;~FZ&{P8O}OmrDB<ShmT_4V}u
zA;7l)Ae|7+&49lTG1LVhqQHuN2;>In(*l1Yz@O)@f<$ltAgg11(uXrHVje*Y3IgT;
z!a0W!&37Q!hajVX?pFbiX>|?Y%2OQDH{9w2@a@l?00;pI{Ykx%f2tE<9J(`%EFfMS
zKnlJOBhUpf!@z;wR$M_Dd^7j}fE+wSLj)Elc6da<gMfu|>;v@2!vQX*!T>_l4fFpL
z;wm!8!)O7(2LD|u-m7D6vsM%4s3hRwDS&{<K41A%F`z6Xy5Ic<`G;M_349U({qAxO
zBgoD6Z#q1_qOgb#^z;;JRrxb;ATRXW(kZ|YfPn>*j*tRYpaZ-(yP$rn4vg#}e{NAe
zPsLmjQVbwpKyn;y27nFb6n5|z0kDjKLm4RS&KCsyxqLf|laPRL4B{{b055}v5q!G4
z<zQZZui>%(cl+=SKu71~AOZeAzpYJ!jMI>j0bhUEzuqH3Wl~t~&ng~&E#LIzO-zgM
z5c}xJ;P0WM0|5yM9TWtC+8{s7@qi$&XE=X|)u9ezKuZ3&=<f3V_#0>N;2pmZA=vNc
zbiO2UoeBY0dm`;%(Sf**_=SJWJO1I{e;U8kQ-4y=e|(8u>^Q$$b1&%oe-FYu1`dDx
z5ID}gj2x)vgpOJQ-|WjcpXn~Ci4Ns|ua<@c9JCaKI{SZJVi;Pur{Jzk0fsobe??RJ
zZcF+?!NUj&z$93&SC)Y9p#%GWj|a;q=W*VA^Bi)&Zi0@7ALf#*A;gOp>60Oj$YBKv
z2>wLd369*11OgD`;^Y;3_GrLCKoBHg=s^2f&_Fqf6h->TsHs38j_}L$#0&gL++M{B
z9G*XTymgm*Uwc|%LO%s@BFe}T;+f%gs_@xyR4Ev~A8TMIV;rl~OPwFUWeWDa^8e~e
zoznUJ8v?jgq;71>GjtV?xD;~CIN{grDmu{~$Tk`tSnj9k7A#C^6VH$CU+N0KZA&lZ
z%!J7NkVjtzoa+Uezi!j9?zLMuoqBpmCtw~c;(fSTzN@&S%|G#FYcT85qpuM1QR5%S
zv&J;`=)VUNuM#-r=G;(+=U@&CH{Xc=92_{rQhA?~MY$8xDJ;m<N(cs=6gh4^zISyh
zBN@6<572^1d7nywt9}t@0ZwAcH(ineMd^*vj>(!y@W$U{_^&0f))<WqpxFw<c}E_n
zc%fHp3;R~GZm!;8K-ryf7Qs2}n^l5=hB%{Ga+0Uky`!(0G7SjF&#tG(B+=gGdhECC
z+HSA2+U`NfMwg_aNByv#J3J}Gd$)5iY5d%LR2x8;iVD3+S1|4Ke6{{&Ulst5wh35^
z^;~RZ{Q^|a=5zwn`QLZwIz>VHMm#j;q}>s|Bo{GZ^0+XTZAk8X`RFYq9d25yY700E
zV6n>@tka@&FvlSDBS_XHeLVk_lggCKympc(H;Hrm0`kZm1)4FDGT^iEI5w>tnqpn3
z5MXo9+5GC%Fc^18w-SG1TA|cpO4~4_ktQ5q8nN>Xl-z&+Rvp`=u{bGr?LLRO_gOWS
zrizEv8o-yh*|_nV2(iGbK>yKgCbRhY!WL6wPfr;Avh*ygqZ6G7nF3fnIq5RN_;O`9
z_pZ#8+k|lY%z^1%n2xyd>Fnap`lSKv*qYYK^YZl@RU#V|bmm7W`ICJK%W~IWK3BpV
z9YzHB-LtVR+tds}msHPKvjaI{S8I|D>1@9&WKJ4;LbtwxnV#QbdyP(Hky=v9hU~3#
zr3JC*HfQ(Gbe7g>(U%U|t!g;boFCz>M0?T}jG=hcmy9yOzsitF_a`04!9^DwLjrH{
zS%d&pSF+G9@DvivsU^J-CBrp8m8_|UnVZle(TQyf)zryV+4Ol4q1gE(zZ4}C=x3C>
zg+E+|1A32PNLogRu)p;W$x7Slc>gF+-*pG2+CvV;p7@Ci$xX?>hQbi9U!Q(JA+;ov
zsr;C3-dxpt&R=1d6c4}2=9KVJ2%|j}Q{silYmdGNBiZvhRw7yPIFt38^(K13_DE7Z
ze2|S;{7Xx^y2zx<#U}moR)gR3JP>;o8}VLVtQ|PdCni(Lcdaia*@6L2(LmY!<3YSZ
znU{Det)te>g=FDCQrmUPtbO=7-QTU!OB9vy(eH1W-;gDnl?>zsoE@X?Qy$vB6>7l4
zckmIIaOUa%;*2#6$;g^pcSx0aH#Oe1q&&MJ8leWW{-VMjJ`Xos#5S37hP^p-00UVU
zyi5$2UHV1uvmLxB0=nvMAF<bX*kxCZHoEh|El&3~507@KkP>*Yk!<(r%-H=5FqeEd
z&--^Xl%6wOc}S0c3tc{=7i)FUF*KF>`BUj#;fvaPITUK33NKj2jjn31q|@Wg6>83j
zIrcm6ru?xZ3G(ykQQCU?vtQf2>=wTMCdrE4QlM-l9;fzbK!>$zNo1QZ$%5&r;MJb8
z7IOD_Z!{nk_fA~ZB=~H=RJuD1Gej&W{Og_v&$r_@jhR|Fu1TsWydm%bV%+KAwShH?
zgG7;5rT(rJQ~L=h@6725{WSZWW+OUnP-Uef!>V6=%WlrTn=+EyciK*S_W<AC?9IIK
zJ#fRXvq7Iq<DpF{2Tp4%(<mE)_`+pLTJ?bN2xv!p#dej;2V{%U4~9US=0HE8TYE#t
zQ~zz6pd13guaf<)?D`fcC3M-EknTjIk(+weT2*2f;Y7>kj};9emZZ7TH9$ODw*Ai?
z{Pq#MVY=+K5*LaG%c?&r!!{i3&JA=dJzk6|*$bdAhPN5a<+OKoxtWzzd@(^xDK^`f
z#8bB^&cSAFS7B$ydyFdUMJ}Dn)i`>D-kk4Z+@-Ba19^HAM*H}AK$lPF()%jU#5vvj
zkjd(DKkE3Vm;Dhs8YydoGxqfz*KsXN$j*J)QMo8$mu4+?Ox4`_SuO<`<yf}~#;@#=
zGL~XfWuK{1t3~Q4N#ZrN08*XeP*VJ$3Cq=p^6J4im6^K#7qFj6q&o`To3<PBIu>Bm
zeLeA`WFLJUOFkng?#IPm%?^BYne%9_Ko&i7Fl~hDanWS-yPfe7IqRgb)OM?U+<P4p
zK18KoVGPoLB)&CAS*r$B6baDp;!JMQog|;Z4rzoSQLdNs@5GxP|35uLsd2kTU6+rs
zfI%^Q++Pc4kA8hK{VtMK8Lv&<*`-djbTwH=qr{WmGwqa0aP+u~H@A{TcRK#5#SY^O
zRVo5`%lhj>!}T$I#K#$saBQYFv#sUOaFI9J^{VV<KaetIdb5A@z@y&xgL}|&9`@w(
znduEjq&`4BatHr@Av;Y4lr;$%f6elSRAftBlU2CvaB_QMy4b>T%DjN_Lw_(}btQTH
z0_ORF#%bYJ7-9Ix@VB&O@Rt}mr}25s@d|94zzG0GJ1xp2rI8<|Z1_4lX@)6cq<pM#
zxC_h1=;Py@Sr^gcOEf=uexX#_v+i4~aTz|ES@8*s#1FN`r0&0p>fZb|#%qd%s5K%^
zR>zBwd<>k$dF*!ggAaeyF)1;fJ71<H?&QrzZj6cI6=<xUN5*A|uenFdMnt`E4^LW;
z?s{0HF(0pQ=!HArhU;g83%cI%mLi#jXo_=icSC(MF#(HP2nW!tx|JZD#|qyMOm`ly
zR95e_s_TndGmg%AmvYsFLn0iGEk+|j&h*@XYHw|%r|3(gQx=5iIYwJ1?(AOt+CaiN
zSnVvdmxt6YKgQonMvZ(pqHf1crq^(qGhse|$*F<Tp?|t-Lh>P2^!g?1iQ#y>@~=yZ
z!eHyM;nTD1*8(ru$T*aGMXb!XGquS}<cTMrZX6t5x$I09dzp1|ex58(8<hs4WG^WK
zPQY>gcjj8HSv5~pd)m{ZtGN;;_*~S*MinjiK(#&0UXmQX8o9Wxi$$!VqlA{=%v@5F
zGeg#XzJ;xmk8Xf%*2(uDNC-R|7Flz(Hr4PW)JH&R)@V$-`+02`aWdFV4eH@E<DIGZ
z5U44Zk)+keYHWs{=X7hjDgO~2_sAlTa&)uUO9pgeOKdC&p)yoh<-@l+8J@f&^_%#O
zHaoRk_x4^4=Azm?z;aJdnG&8~I!bFU#)USSJ*MJJopDf(q+~+PalaidoX3jxdrZ|x
zQtQ}k9zH8<e$?59_4APbeGH#<s^#U(c!dtkgsCp7Pr4$JWk@ia;c?_`HeCHXHswx*
zyq!@Su2Oo`OYC%2>jZyqQotgepR9n2<24Pes>0rl)XVLlObAaWr!!}gT+5Z~b5sp_
zHmBHz7MXhG`FYgLMl`y*of<9!*V5bP=NGuu;xU*FFB5DZ9kl&LxX!t+VPBG5&>D_Z
zXLMes&3lK09sA<4BNqo<Js-wSBHBr^3C6;eSe~z-XE4D<35{}28u1k4Gm^KTyMrx|
zqt@~l_m#Q?y&$9*KDpR-m<%Q~?++J9EWqm4N@TC(LK_3y=S;1`AJ!g7U&>~Kfl%sC
zP|3b&aUIDQ8Q+*QdmF=UWu6l5n}(7wW7qiHSIbb;&Tf7s@8_YxP_ML}T;4F+SA|yJ
zs=M>-S&J9tJz8$Sxf)aaA%>P=7=#-N=W(R^(R`;(EKy7<kusHMy9K59^|1D~d#jGR
zX)B{^LVJX#SQ0_rDdbY4e@j6A<hhDE+gUK$Qqgh&rFCs9b|OXEbYC(bmo2(j()23n
z#S5k?=qL6%AHfCIT5~-+#TH23HAD2M7DnPa^lVNKEmQ}YsXke!gft$0?>1wtT<K)x
zR{{UL-M3IdI}7;)cvGuIPZ}Aq;pD;{owKEAx`^h6ANo-qzDUJtpO_xD6>g=a$R_JR
zSzhDIMo<wEk;AIL#T>WCOu}O8Nbjil^7W>M9~|EeS!&!z_MprJy(lH&4bigWdwI%{
z`<x9w*`xnz{vni<IUEt5JlcDj1sru@=y5!sNy_jWEVYUORVU*1-}YmY1!Y6Gc6Yfv
zBf%Xxv;3;G3I@){wjzVLvvEbpo0$%Oc+<EUZvP0w#0$+)`H=Tnyhoo)RjCVcuia#e
zNE|yHWS#O@FXCPk712Uz#(JqhpPg|HJK+Zx^w4P)zg%Tfx^>a}qoPwk-KCn4THBHq
z3psa&DTrrE&_i1lCNs`Czbf;MnQ76TkNpf&en8o)>et6jMT?J>__u82hY8ICv6Zl-
z(Bj-m6A6x=!|Yc~s0i|&zT!0)bvNA<ZPnT_Z3Cn>`}%!tSd1!FQGt%AiNlLLPp+an
z>1>%%IsfJvvbq-*wL?!BXkR?v7|5Y!pr4L=0<T%hd@{EmAC+MEEw0$|t_OpRBD>Nx
zH4T?MBj)ZU$$3LHmd&WHpU;eF3xm4Jk&j$3vi!U<e@$*MPpYTXsVXv=XD$lAEEa#*
zuCsBpN%Uz4??3fBO4?EaJpjW&1<P~jh5%qT+tJt+CsH{@HZK3%=fQ(f-ap89@Z0<|
z8`Xj7-#|IKmw73TtDH)3Cl+IL7Sf#fGl>MW5!|fK#|yp3B9>qp)06n#;LPyLahed$
zf3H8Y3{W<!mtkRao&Na~EhFlht(m<i(Q5VMr-M|ip5u7#n0*tzL3I!ujyVZ6!6*KG
zDVha#k@fZct)HWA2?XMKcel=fW-m({DL&uj8#wQjSa;{U^Bt^Dj)ymT>?b%ZbhKc}
z_NEMDDJ9tbSPqtLdY(IeSgDP?>`s|$s!9YwVXf^AVuTOn7W1gPv%kA{ofqD6R6od_
zOyOJXtXx^tdLgIou3AmhD<lhNM%8a(lm&WLu#d+ev|?SOA@FoUu6$IxP&rDh{;7z(
zue$RB{hw7g4pk8F8fOMCzh>hfHtM|Y^XsKdH?q9wZ425)<VpJ1S-Mgx^FvgvNIqPd
z8^=eQPz8fygM_y^FUt}Oo=M0}nuKoOZpYb4<ghO;Em|~aahYKJd|>=_=s<clK?M5v
zHJfNIQF%+RQOxnB4&97<o}93Spj-qcZ4!S^8e2I@igJ{fp)KR!1@b$jIG(gkO?^|M
zPPb5$!Rcmn+bGhoK$+X{_2ChH3jOPzA#jI{9R6>LXQkSPW9;Jc*$MkFk{fEHsUdAH
ziCyS+u+~GFi0~Xr52Yl};BWRj&~gp7_R&NMZ}Yz8)%`>-!+#HMuGquHxCnRvna6!x
zBSv0hwNwQR41w%~D=C_{G*&JYIz8y@Mho-v@=%Ya{gVGpEXE_6Y~0P<w6<UX<S5`L
z;RcB4sK>aY<@qud=$5G5T**9kPtrnSZCwD6I?I%_){|mv{__)Xd*+502bbsIPk!?H
zZOaquE-q0?U2!V&)Dcf@bYE_~8An^ovdIJWbqDGdI0c((I5h*Vj5vMyXV>FU<Mu81
zMG74G-;vih_M=!2eeNu}REF#fngeH3$L8hS%8PZb8HeSTCf{C$IJEf-n%PsBl>4ZA
z+%*pK_xEd(tJR{7BgCKW(|B};ctuBgL+e^`!e~oId$R{HEw{%FB2+@6ZCnw~k@rs|
z>`kN<nsj~|NJ)b=G8e1chl+`LdYm`ddF4%F&UyAq?^qNL&bp<K_61Qi?Tqq^Oj%Ws
zN(1mXO|V__t`-}ET+R=#8h*KJU+t7uwCi=q+{(s(#lEEkc_BEPzU?K3?u__Z&TI#b
z`O`ic8y&^)<|kO$lRi!6E}w#~J(N>n{6K%>RfFv2w_DtN=0%Faq_$R@W`x0NO0;;O
zOdyHe6(CF~fYvlF<_6+rV{>jU>^5rrG`;-B%`6MDL*FN8U(J^bCDq1jN@QPE-rai?
z*_IBWG(&0^HEOL;GpQ!1`t~(E$<+w_Dir0rE`!A}ouXEY<$*h8lfCe(;#X0)G6%4B
zWS31oPjQFw!|3@AjZd6}1dF=jXcejbx0BGThjj~vsEzMT=U-`281ifP<5;Q#@og^C
z1T3Nf5A%&vS@<dyDYYrPZrJi7@jdBW4s}+ha9Qy$-+GZ{rERIph*JD7XO4(68&?!r
zX%st5U=k*VENXVVlySIcr+i$XAuznm_1`o!UF(o72?t0EBaXG<%_&FHBchveQ;WSl
zW9-^PSoKbTme_d-L!nbP+`Japvn`LArrBq=dwlKKYv;Egchg(C9k{*n>{&I*JKBJK
zuk10w3PCTVwk~rzb*(d>%=Ob(J^RJ77UGK$nAS;6S0>3znvO{_X^Zv#?$;y}h8>Tp
zVy~IXz|f;7?yAsJx>CyR?4Fq<s6RijvF*KI6mLplG`G4Yo;Grk@3MaWFiVE9zM-#i
zcydlrb$Lm`3)L!G6h6W1p{1|cp&8sCb2AZ^$>b{WHXU<z!R_L45wUwLU08D$EHz_J
z{bjvKUd%3!CGTTL%UbW&;iS)e#-pMDzilmQ0<pa02G^d9zkwxC{Pety*S9^=2c({q
z*YO8{0Y+%7`_outKSoKb?ude8Otiwg2~ajvRJrE}e5wwS*8Mg4=Pwr2H%%0JsU)E8
z;ysds0EXqndekdAid~70!wGVG0ncsS=h5~o&0_`Rg33%pP4Z@KpH`;rivY&T+ChWo
z9wle)QCa!3HbzlW3n89Y`ZzrnZTGeraM?5D9@DoYd&}OTnR)xnyY7v^8ym}~0mH?U
zZSSm~G-L6%s`mF*AY_3~xi?7)Y*tv2%+Lt+nP-l4kfC@XM{KQL5$OC&0(?5CRjNU(
zd*!aUQWU~$#pJnZZ)Guh6Vnzv$_fD%{mjgS`d#<!I~krY&ATU|;`fKjUsTcu8L_1Y
z{|3&fhXJdlHsdVA#Kb>H^dQLVJRdX9elhj0;1jEWU&%%{sUFhJ`z^(`nS-YN?(mG+
zII6k)88d|#J6}5H>aRMZjE+QQn)1#HYE*q_JR}OwaW-j%4J?~pc=OnCtW8qS{Einn
zRWQDi!`N&>=s=xFdqQj1qo>;z4UPJXJPoVD6U<L&LJo|;UyRK!r42G$TzEq_P=7#9
zxNXbc{&|_R;=U4Mom6F+oaEa_?jenyv#ltlSaa{MA+cKCTv{(w%kaEag+iR0C_K9M
z5+%Z`N<NYYl?Rd7nR{}OBRj7We1&kOH!L1pHAF8s(41tkJEK;LZ{eu;Q=ss32MZ`l
zfGMc)?G*07HXTp?2#i|@&@M*N-+g<an&^M;48JY^Esif%jiO~A6X-JhgYGQh{piI0
z_s-Ya{*R!G%e<_p#s|Op>tfkn0Kzt$)^xuK?N+-{$F>axBb#(l;hnL4+QT2(i<r4=
zth-LmDLF;QWZ`Kdlccl?UEfKa`D_8FGc^Kl;*Ioq1#UWo;6=f4%NiLeQ@&fW-1V4>
zsfCiQsw#DNpEJmsUSjMCcCP05HdOB!7328vyduMkwDWAgbOastJ<RVL5-wVNU(}la
z(%)hHnZvAh3EtubfB!|KiRm9FgdIov=Jg1ETObBi9r(C=ux_g7GqJ3>Kb4XF6GpZO
zhvWqL?zht)BCft4)<-YHdno8XDqBGNc4fX>2c!sj`gyfuq+~-TWc)rB8OytZ%$mXe
zO?|L8iit17!kI)m%8V7cqfpx{#R}zSv-41$QgEo|57TK`s^V4aIcyljhx8e`=NmfZ
z{MF`OLo|je-ouP;MJ?-9T!XZ_-ArIDzX39xITzfkuGV+ER#T1<B?u6F-`!+W&9BXl
z0IceZHhy`%*8s)%xbVLdR`^!TdC~V3Ik)fnd=w{)GZl((HLM!LTpTZJjW6Ij(NuC|
ze%1$|pyJ83#V%pjl%*uvqsM-o{PK@Zb)l6-w4*Ci<{vdhn>Cs;x|RS5SurM-k}pnn
zd+S6>wLLld!)D%SuBZ1}aaT@vZmuxQj~kxXF^e+x3<)yBTjMQuDWuvKn}k401}5Ac
zQd8ZUb{cD|t7pJ_o%Z&{hDLmr3=4x#ep6xonE!)WWBoswwf|<U{zsrO5wI~ba{N#B
zznV361`fvm+exfg7f_Xy|IFGB0j?6T#M%4~p_KbD0}R6m42&?0ummwFQK<{1R5{Uq
zTniyVMM;P&br#OY_rH2IzOSxVKJR>I-T#NMbLtU>>7s1gwr$(CZQHhO+qV6*ZQHhO
z_q@p@bM+<j6RK{ia@N^vvy*2BHWox7ZzC2#h7*DY3K2B{u<F7>7eEMzh{#BYh@?a$
zFM&io2Ylm_AF>1!_!Kx~T>CRDBOo9=l9j;;?IO4yF!b~4!viS50HmaVN=-yX0Emc+
znEa9rA~pfICg3ka7Ept<gNF<HL~yL^>rsF#kDz)s&+XL%;4#nxP!JQ7-gDy?96|~T
z3=m)lpoh4GdhA|9fO!JM92ii5M1G1x;(4l*EJ~|stncmVLmXcagNV0NJVF6{7fPrH
zAP$2E-3N03`o+S`1A7Yn#*m2&Lozsqetv^u4CN;N5`cg>Am0ZB5i;awqv&TKg8;T~
z;2PA_0?)q%_5TuGeTyOhd~@OeC_uc|JNY;HqXrH7kq7r1w3nkpgar-i0L&`DLkQ?A
zi$h=rpa29I$n`}Kz~jHm@{PlTyZ~_SBKWnz0a;XG012=o|7OmC#{_s0`iA%v;PsiI
z{HmUwX0RH_SvkndL%>`JKUed+QUC$<a=NG7=TEtY5BwOO_o>e+WF$Y;Gv@g8gvc5y
z*y~5Me&Wx{Mx5_4qC+4N5E2nnky1hicmNXs8mJ%WV`mg}6Z8r0*Wb^;3|P>&qxXN9
z!ALMHV>&&Fhr9xQ2m(li(34>N<oENdiwF?(LqdrG;Ska(U@Y`i?}ZS~`q$du;se40
zXhd(1!$Tl^ynL-?-^32)>@&bG_g(bWYiMi<&Gidp|ILm60ja7U00OiWRB#9=sE8?n
z0HCBJfJ;b50tNkPjst-FERFk<SP$hK0A%)k_Hb78dB1)U1Lg6<9e{o}r{R5zbs_-F
z^`+Vo6A=RSdIJ2-dHv-+{w06ZQ~gyh{&f+{>chwHam?w5{DFge3h^rVW3Vs2i0-Zn
zp!*qvt@|s_5c+$nqaT92ynNfMU_*8`218Ac9O;8YzAS`*3FN>ONHkpOM}1J+eNI{9
zDPRl(+yei+xB@}|iHiI}cQInD|C~Y}IulRrAsmnL|7|M)gN1SbFlAMQkN|`T1Sp8G
zulY150TcwX$GHT7e4{Z5Xh498@0I}AjY5NQ3?Yj2-KuC|5nLZ#U3`22pMktQJOUK@
zG5!LH{sM~#0VVV$`d`DpiW1nxylWG)&;0cK-grTf005C?<pl(v5banIYs*!sV*G%r
zhoSH}*JYSLLKT%G*7lm4yjymwZX7&C5T~}l-;JB6v3S6x_BH4gv1eaC81M2i+VH`5
zHu&4HG%d=ou*z(GReC6w)`e##NY0CRvYsz;kJC)!c0o4(;F<Tf)w5&@60Su0x2Xx}
z&Q;gr7n7^Os8^HRA?tTzp~|qxHb8naj%w~_Fw4!wg`=2|?H6xDx3ju=JZCc9EwNtp
zjD3p?Xh#~6T2zBBlq@~;Z#W?<=EH??!>Bz^uZO4Li?`2LSzo-JlP#2zI*S_HNY3i5
zX-Tb$lNn05@7dTvYWX4o1Kf`l{8K2s1Yc>PRc)8JB)AY~<`~FJ)6rOf|Lc|=9dcsT
zE&qC#Szpj$OCz&@>vK|?gAYLCfrYGW(;|EVCHX!uTID5GoGU>Z_Hwc??eKYlvQbTk
zyc48|hz_s&p51#;@QWTv>I^TGEvuxDxw0Fn4tv?=y_PiCQ0A*dg4XB%Gv6tJq%SL$
z*J&p^oRXz?tN2)<o>rCeLOUJ1)M(92{~CpL0y>W3utnrVnH+)6cwdwBL;=-$w@ihG
zZ&eq{nY)b?x0903E&5Snb%lqL7n1MoF`I*;{EN$OI{2He|JL5qm_~(dFnp@=R>lSn
z9)8LT{jwEu&mp;dc)^1ti1o5Fy2kOXe+l5+o-&?@Gxt`HGh62OQQZAj+A|8vX-HYU
z3Or#JAlX&d`U~ov5IrGZ-;YV(-ee)(sx-ez39VQ&WaaI4hU?uFMWM1Jey6*qmHQUf
z_?#q$n}zn$!ws&IQrui;kE6eCJGW&vFat6+A5)P*thHM;W^uP3yPPg^QQpQWl?Hne
zE*DA&SF56)^N?FDh$fxg2%<HFT?Gr#8tKl3mqT&5j&wst2Z|NjRO|TC+)+W9uwi*1
zg9B>;j=vM@sS(?mpy&?nC#^KkK|`v8bihtQn<5(H|L$zAz(B0<@c#FbHr==@(VD-X
zj_)T5-HlF0CVJ^JJ+sL*;2FCsv$7}E#|0)sg@@1hl?2*GrtVZb1!qg;u^!iK!;2d@
z)$6B=->g{kK++5J`7+Y_LX6t=ET1l3(&)R2IoPJg5(9R!dxu?1K3SU7_$KxE++EW(
zS5Y#`RzY!gIJw0pR|IOcNtzOf*=v$}lK`(?ro9tAdr)YU`WJ+moiTzjj6Z|TV>i!@
zAWS>?(li=Ly0eAs!#LMiB73T})xP?#$P;9N4?TvpZpd&pmn_mGl-B<bBZLmHd6cX8
z8)-z8YHrVV^Ydbi;%07CThY&Jy?(5)M|mRFl9>7E@1UD{vH)r8YGICgyA2Mzc0P(0
zC(m>NkXHw6THq{c5MLFNiuNbIHWe>C?^Pe%Jj5?;@oRF6ID+Sxwv~Em{<JY$a^YVu
zD?Em+Z?om0j@3@<#w-T{M_1fzKA~7bwQPi1tM)I5zw#C3?GWt+rFmsPm#bb!3HJ+j
zfNI8ddqzd4%ob-~>~c9B8$Ed(kv(!3|J7p;++|lA+Y3n1xQ9#C8cHfprkjxFZ_dkl
z7Yc_d4GV+f_+N$S@z0RL$H)<SYb^ec6E!){Lv5V9)FXd_iE3jWbxT%uht}>PQ>T!6
zLu8m=uBGPD?y|51lYfQhaYn|So(;a>-^s~1(ty_ssx6-G*wj{VboMsxWFt(!UY2Mh
z7|TZ9?a#(B_^bHY>(Y)^fo8HBs*U*cZ@VC3wUX+ylOtf|(y$bD-g!^)^Q7sBb3n?1
zc~H3U+f*qMnj;fAJ}GjH%z|rH!c<Sa-d5Fq8=Ab@&qVk@%BtKxEglhhs(v|KLVp~S
z`XCdZ-DwtzG&OGKR@<uP1|Kc=X7iC4xemsCmTwI_r2^_WP?W|7Q+5t{Ty><9@G(jX
zzx2$&Y!yjl5XK|in5G#B3&r3g_Uz!tS2kz0-)3@U5?oK};Q?tJUp!Pn2mJ@j%ix!8
z`oRfa>R1hpF6qG&JlyV<afPnMuDth85gSgW>!p3?+r}nom8TEedG5&<s|ZBa-RGR9
zEmh}o-#K1)w7QTrgvm;Dx}kPrGMU;h)1S)gK)F*L?0`7WB_a_~{hQQV&dy54LuKvI
zBh`{Q5C)Q22P#kz_l)q}uZJWIEa4-eH=?+tv*s<nb5@3iO1b{_XTMD`6B|?(s~f$R
zdEoC8rnwXruon(%c4TSffWrWtv<fF^OJ2w2=1hmFEL8S=WY96)CR9~DMCsr7G#GKm
z_0o#h?J?bn#a5NP>1$fgA)OD(hlj<Rgf8_cM*Y_P>^U&K%&7WN;>~-~<Jz8lmGJ%B
z`pVjt?x=I%GIYy5BaWfpMSiMbbQE(MXYGMT8*S4EH9xGqiaJiL*gsyIs}$EV>gIy^
zo16_IYKXH!AH<+xXh?~3J#K*X$3CZaXus}j@tM3C{+OB8%chi&nQK8g`mu{%wq=n?
z#Ju}y`{-PTZ`d>So%A~g6(th#qL=wX#XHWuvJJAu>ijjX@meY?`_|=Jw3oVemL^m5
z3QXa(vQDOh*&$y#7e~D3-X@1e9~G|Vi(Q+8<2+aj`1Dp~IQQtUi9z8rrjwDHazEw0
z$)ma&y6WwS7Zqz;VyTL^5C!{!@KYn*vk;{7^^iCBb;gQgj5TsvdYh>voX3rVe&1(N
z+gMc-)@!wS`ItuuW@ErUs)@`zdm&$fIc`DeoaLn`;^MtlPg$Xr&A#&Zj%IXTG~-ge
zpf#!|%uG2b5?P5+x+!L-h(2U;2MGv;yPd1ikSQw}g1)8#S7K656Jw?#>imsQ|NJT5
zX(eUP)a=n4GwlKgKR<mQw6_3Hap6+w!gezWHvx5jm}TzC|3+{FUKbt3{WQ(w7FqR_
zaeSr<N@PP+b(^hy?H#0SL=}^PrGLptthPLh_WI7X^Zr&4gp*s`)6<^`ra~T?4v^Y=
z^<qF>;YG@rsGm%HI$#E0X~yqZ&Q8Za7xL2q2^K7)bl3+1iu>T6)G@*yNeGLL@+%)V
zH@rimp$Xf+?x8C=#V2Bof@~yJ^Itw*Io^cX(<Y`CUy5{N7dqx4JWEVK)ZcK#;~Nwo
zlKO9&H|Rwi-9g)Km;i{3;MR-TWx&G?;UA~Kn`v#QL@@2?PViN8&Q~e7WS0v)V)W0p
z^+sJxr!Q9K(+(a2SMF;HuJY$)kheO-6{jaxi7=Xu3eYkN>c6k#Ry+$lA5)#rC=z>X
z<_GX#_<W0caae@D9SoAm2@OWkW8a~@oyq{(PiKubFFQe%DUk5eN;A)0qlMfo3T1Ec
z_G7LL<7JkL+H|-&V!c0ws^OUJu@4BgcA<taEqb?hBrOK?r4cM2i}FhjWLL|-Wnr+3
zCE3_~9l9&}C*{G$Lf<)oOtm{<MjwJqz)~2-zNcqlnQtK5Jo8ZBv?4?qwl|fKH(O;~
zJzzpWu_@}3N5f`%^2zMsJpOnPbjWmY$K@j=Oms_@!^`ocD&r*aORE@BG^C9RV9*fg
zP35Z@b1>c<tqCY+id@)?5@*lS%o(H<3mNB|<C&Z8b>}7_uQrHX8-wM=IvLPih#uVe
zVjoe#a+})Q5cwS?Ng)#_f*+m=v(PHsb!wLwBpBXuhO!s-qOa?1>e-S$(M~xacwZPk
zl&+_)8;Ws|V~dM@_#PQz&jz0ViX-RS=k4fU9R3?mu2^hydR9M39p%cZowx#`G-vJH
zm0FGl<XlhI{X6JgI77{(By)!;Z#`H3i)-=Qx>NA4f()})k`V}Aox*}Em*Fj^hXux!
zI#D_hU8xepM9049GR7bL(N^*C-J>p&^;)(q-PN=ATX)vl7UJiuCBQJ%U=m;aj+lQb
zikc1ggFGsv;y7(XEFgYqYD)Fn*;QRvIe>p{SZnWLbzOKcI}b!d13$VrUO&s4b&pNK
zMx8^kp)IBOXi6qSKnzm{U~?YIoJ)+z8&Vgi%Yc^aWTL^%H*EXJg0%vCv6Ye3*noD7
zdooHiy{fVYcCZs1myAmZ6s!BRYbRH10Z}Yx8_Nu?qG;gPG$XKu)3ooy4!~KQzyEmJ
zkt{vVbDi`WWLaVfLa?fTiN`vSX<JRpUxG;O3V*fauhapE?i!b`VPB&4SKaaOw)$te
z6jv%Wq=ZI#fRz_)5z{+TpqPR$d1NBKa`bnpi?hnoa`2D5KmqPgAtfYFx6(X#e088V
z?6nj%08f820CdUB_q?B<Ih#F(QYcKBEI;VhJWaWz(c~N>ShPsV!3F3~scsg0YSBCY
zT4Az(>6evBt)DEOdb8#Ab%dcLZJ^mkTR?f;k~5jpE++G*?!Rkyw<Fmw8#oZZC=#j0
z6Zw3>i{r^kTMX7hF@Iv+I$ROrEkQ750*cYvQu@f+M6nV#eIA1!p{mIc8f`4W&p}1r
z#}(%i?Ey`~9Bkg(QttAqc)n8;QbCJ0cKCz~;}|#HaZ`)9ERB%nBEhy})TDEV*~0u_
zB)(irb-22CmLr=M`zKxo90)rHeC-6IvKl^Ge<nPsF7I@A)^PYqq#RR^&#fU!V}Z%p
zkWtbRVY_pKqCcH^j2F1gPUPyrihP<k^w((<XXqH0Ge~REC}{=OJMW`D2QZiYXvj3e
zn{LSSurPn~5=!(;={@L9!yDxLz2V)ZUly|Z=6`5}PXp6zHBcL`pyJ0GAKKy^qy3)z
zWioF0AJV>Z&3QM*&g1whYlTU13HQK4TjpqX`LMXTg5!jHvcu4fqI08GD-y}YquTF|
z#4xg8>q%{mR|eSF&FOq!2qDC+ohF|vUvB80bbF`Jw17@Ex8BbZDQ9QYVFN1pJpU=M
zc1JaoJZVzauFY1lUw)J2$cu3*PCoD7Jset|cXHk$8%)<W-JL=g8sT@?8bH!t;Knle
z9TKV*U9@C>ZJ&wa32pJsZ;gQw%IXn9b(TfQ^J@1`vt4vidN=f5<0HGw&DQPSzl3OG
zt-%N@T4<w6TG8*tQf88s@<sh5>X4;TkV6L4@)3=DBI3|o&ts8xFx23U$JUz~E$Oov
z$KBH}W_Bsbk$iNLikM+Bco!e1jY!N6svJ~^?0mDD`8j){)amGDx;o<s-qjSod~A9O
z2&?jl62%=;BWz6jxtf=_e?+A_O9$9$nh1B|NJY)vSts#!siiKs8I>VcalyM#ucmK0
ziqyr6Pu`5X9+Wdo4m>uo0UnHiUJ5zWIGS@G_Vm_y(Jwzgyy9|&Phzhv3(=EJumKu@
zzY3+D#qWIuR9%X0xZAHtu}x!t#uY-4N5@v?Nor7iu^Txdek^eoGpnZU3nx%udZ>Hg
zmBKbyyV!PVZI)1DK>SBK`xZj+1M;|9Z@zCdIh1K|r)`VW?umK97HfXNW3PB*(-KdY
zzvY{~JAAs}xgY_Pf{^X@6b;Ga5Q}Y!rJt_ppR0x0mNh`ZDro9jVdfbU-jSs7R&Y`5
zcQ#0^3D~+unO?T6XyyLi+pc9D+H*9FBCoR)&u4M8PcOzRnQnrw#=?YDl5p~FP&l+<
zmnp5B@h$jMsH_<p@Pe}|>S<~kV^%mJ=I5wUGZg^J;nQB$quQDQNVkNZY!<v5S+m|{
z=Xmm*%y}%fcsQ5URH2l<oR0BY;wYM9sptI?lj?(uoWK^ZzwsXm*j%>xEXhPq`zR5f
zZ_pU(Hr3EAW^)i3Bu^I}2a(5>up>d@#8%fM_VzPp%isqZXwx&y!^+Bb4f!&tTU>7B
z-5OOZUY-NzuaK4;!`0KyDbd&DaRa?slJUVnPhcA-{$W69_mDb5S#Tx0iI=BHarT~j
z0Q7caJ@W<Pt`Jd4mnqkpZ=`+dxtr2(IRw>{Xv=G_8?&m+yfaoBIs}jGq_&H<Y!q6>
ztAl6;)^Hf)@KZpZrEU(iaM3Kr^|x4#xK?*DGmJgVrc&JiK*Obt=5Mnb#jzGg@qRVO
zfl_?s1Y!8z&o6@VC%UvHjXr8sdGf)vtdri;^8HKIsdrJmb56BE%BQ8fgT`)mzt|%A
zPlX^1K;EJ>%}WD%1b2OZM##(Vnl2Os2Ck2u*pZ1TY2Eu86j=C7_7fajDs3~;7^*-q
z(azCROn$%}Wh8&@c|O_v+lojC9IoY@IDlxkkiS4wjm(T8?05EyC)AqY5nn!ZP`EW+
z-aO4~lbVK!8K{YT4V;dc^>iy1ERQ`Ft_DJt$JtKrG-_`3#T7(nz@{#yZ{%~4i*f28
z32vImELC6KMw5?Co|}sfk*)19e_78@RNZnTlHN&*?7GZhUNW-D{|qIiyP?pqICpup
zapLcK(cqb$a9#^yIo^DYn1(?~t{(KJaf58yYNj9fa%7ySy-^g3GIYr!e7dSTxGX_S
z<SalX4}76!;J5pOoKqD!{7)`Q-AK*mVLA$2l8gTEY=&IWB#Zu-YUl@A^fYbY#?^Y7
z>SCpm`9M3EnG0pV!3;KGTZuK2$7pgLF8rEuR=Ow?G9f@%muYp#E^68$(<r-k)V;n!
zt=TT#@PvJY`aKj5@^@cPeT~H3QRUvi!@q8>Me2MHpW;p8_xIHqn!D%#p~+q+p5`tO
z>W#56rBQc8_vO>dZ`14FGu+cz?WThD>Hwae(ao@YRXj+eN;^*Sd=($A&{aFL^C_e5
z-MqW=sX`IqG3^mIJr$qpEQD;17ik--o-v*;lHcMFopK7M6+MK?mipE9<>GS0+Z8mP
z)j4P=pWM19?Z-fjlf_X)f?KT~)jE!5tJ4dS;^fo51=qL=uidTX`qKh}MTGpufSmBE
zuw}wclJK~-<{P&+C*w@TF`A!Aj4TJO1Uv}#5lPCFDs&R2evfr=@#&kcH|+$k7r78c
zfh{$howR?r418KS+X3cq?a+^I_!@T}wlQee%hxtw29f(|Ylz_!3mq7)Z!q>Js+6!t
zvalzK30pjn9Hy^dJ{ok<AcW5NpXA;%OBo*@Jz6C+*DD6WXaj&g^eu&J5au@J1R5+!
zert#ncY~fkQaFZ@!)dR-vsS}31MBwonsFJ4iB0>zzO`{I;ZViSEUj8dKe9Wd9!L9c
z?NChK!{deVirTA(^bybF?)boU#_fZV6WWO)&PMT!AnJN8N%>0+QLXY%<I|F1ZI?P1
zcV;w<0G}uy!%Hn#B>3AT?HH$*&u&kdyT3g9KaygClb!x1oTJeuczIEx@W4wK?0(d`
z_~V$aw~CJFhUi^)kfR72If+-JKLXxYYY~tOR&*1~k?@gnsp58bQJ81O*t?6=0kJfF
zD*ebV+Q4DNu+NTTt0E=O{;GALPwKO)1FG2kQdouNtGI9Lw%xf>Ratr^*eD{6XP}yV
zT*N!Lz7X6HB@7uqKqlvmBlxdWlS+xXf6{0ipnRUm#mznO%bSiV4DI6&=m?XRzQYHL
zG3sck4y-kl+&hiIJ>-6rSJI_cZZ$AQi~5qtw2(ev##bb6_f~F<58$F!GZMO`vs4LO
zb>MdIF|g=uXyw{e>7;C6|1Iejn$oGbliWR<$MhtaM@w(o>1drROQdQ?6tKeqp21vH
z)68j2tLIo>xL?DYfn^_bdsGVq9^965;M8lZ%0nZo7o0#)x%R+t=1L>~8HSkGE3kGx
zCMNJD!nz}}*KdwySh9=RAJI6smuikT&R!y;`w0@*@ew!iGSVG4>i&4%naD~~_PK_$
zzh!HDbw7ylel6FiChO+yGAcG|>tF{ZoaQK(+?pClm<(C;KlFFT$y7cjpc&&J$<q%?
zmHyPQK4O*Jbe7WKfwBGzt}pG)>9YD^QQ9W}M1wCAMpRPeUgC9QU8k*zeV1|;P=8iL
zKNJV`Sz^QHRb?MIwa=8i^%lhua9{8yG7TnlvlyX2szpc~+#5cnmNv!)C{ETf%3DXz
z{`Y6-#_JKIZjWZPcNx3sXW-&}j$IEq3N{XlgSRe-yI_NMoW1cnfzi$Qb<BxYipEI^
zB|2bazR!y9>nf>3*I{Hzf?E0cIW`MneE*%!O~&Y*)}v|im?BTcODnZm4Bh?~--gMW
zO5dra&^nC8<|T7_tO(c`m9aCM69t!LjfxqdJjYSzT~SqLWY-cDR{Q7+14TJzi(D#N
z#>~VK-Gx3qrl<*!SL^#|^z+rw|BFULG+WR$AE;uNble5g%k!?sIBAv10pi|WwAoU2
zj@|4uq0(TgcEKY`q>b##(i0^hmkK+Or^CB?Ie03u3#e|hQOekY)-$6{z6O)=_sich
zTzUFG_#pfL;)CpL|1V(2M8L?v&cX7(03jnIC*%Ls1N{F5=&@~}3fWg^bTQ~}hbbu3
z3Iy$4UDl3}2s8pgU1(4!?(Xj88qsJJbfHEgQQm<?>9;*|7o2B*=C`@iBzw(1Jy)7{
zK84<uWfMepkc{9H0yu^=*EKdk0FtJro^EUa*x2OM*x2A)xVdN;X7HcsIJt7*$3URM
zIH3Q1S{Ff0z`T<rGY9fuQVHV#K-IkgxXA-@#RGEH17czO2S!K#=!GMq0R$$eVQnB4
z%s?pwa0C@xxFjQz%c0jd!Cq6&^y2}*pv3_Gb7Imz>E8l4f+s+)4^AT%*nl+ya`KqA
zFf@Z(2+{NtWSsej8Xz<U4&;CY2>f>_HD+)M>fsTskVMRVxUdZX76FBXXzX#M`!3Uf
zs7&Dhl6M*l7XVsp2IqTFvJt4$JJX;b9!TrKv;uMS<ml?+(h1P}l-vT!dXNQ|AWq*~
z8W;A!X8ru(zy{`~f9~!6-TvS}+`lI_W`=-{4ouF^AlzEOHT|yy0aY<b)YHJDFbvKu
z-?kQJ@S!{(*&UgIH#5`uFuzngFa*SPa17p;_xZmCXmkeT7%8Ywt$#bj_wE_x#%XRX
zDIpvj0XT*V&wDSDfw}@{@@Kb@-{$RZaR~0>cm9Ob4aKD^cWgAexdEGT2IlGs-OK!J
z@<h-4Ox6f6{WGIuQxjvb0Gz-AcxY_>dyp+WxdQ%_8~)MxUiU9<K^%ZJd7A(qLA3(%
z{O7w0%<KpNr-x9FFTe65{=_bBZULn0Kwu2un!&gWe<J=!!8O05@=xx7-T*M`dFNwr
z246oPZ^;YDLo|eOuX+B{e^`tPno9bLvY#*isV%fA2>~AfJenPx05mw-Hvq0}bbJ7I
z%fa{l;)>0{e#_(j?o-`cf&l!$zjW`G1b?c{+WaN@@8`l8fPT3%ppGLu2>^!h%dndq
zn>BfS8~^y9e(2Nx__O}v9{o)}|I|p1tnB~N^A7BT|KfM20d4L+;zy8Obq(dM3t&98
zz{dVqS3rMyx+V#v8>)x?tWFHg<gW;BEq+8S(cI|N*y{RD3@TvOflVsG!Zg>vWwZNE
zHv7<L@vb41f&~Wp>NCjc*;@Y=cx^L7ca~Uzd=F3esF=@}CI^18j_qN*w5tB<gf|y}
zj6S@-INCh{%;@OM^3LUl{Drmw>|e`S|ABaZkJteutRMn;I|1^%TthbmwHN&1vB3Wa
zwm-C=(GO1m08V}h=l~QB;17WTRDTKl02E&KBI02IRK9}O0R>+8&ne4B@rNM+tbPUM
zH(dW>-}4hT`Vilg!^)>d4&smBJOOF{fcZpQKjGg|Hb3FNW2b+Ya?d8VKi~fSusra~
z$@6Z0GkQuH{C&Ouo#$6MyaVUEVL^N-A1)<w@G4?ImQ%ppk|VzJ<K*4@hZvs$0vP%a
z{ACq|R!2`QW&CONl_UN^|G5?h3E&YxKZkNggc?NH`bvzs)^c=kPpxJkJUU+f!vmkv
zC}gCn(f67>_U<-QJ1sKmm=?nCifOcMiwb#ieVzR7cryc5Ehb+hg-)H0V@>tE-aaDz
zmI-=81%1Jhqt?l@YZ`_QL`wJLV4alM#$!>GzVegh!pT>E>Cs2tzJQcA^*|NkR_XMz
zl4eegpm=wMTV*w{Jr&l{-bN?+FTH?k6vSQxL#(8-Scw|i=~p<`wYFae3kOeo@S^A!
z1Eb;qy~$Mw8+;%M4W}sK$XmUjIXTulMd`RDKla9!PKvPAc3rPwD$`r1&e^=dv&HT;
zk+42wZJcF+I^wtg5{)!x8L2?DLfU5Vm3z{5-KG*j8%zYLu8_OS(y(2gTIDLm9okkH
zl^4z2lWgr?7e)C9b_7faGd=k*2%nk_-Ye^pyI`kg2ys(ymHfa~s*h~Sbo-$NdImqg
zX-=!l6f%46gmGWG>0%LyLjMrn0)Bqr=w#Fps9gQ8F;JEtEFbkk-^<&P_c-6(RdSx5
zYfXep>`2_RmX+qC1pRaerx}=^U8`h8RMM-1+o$ij^5NaCgykH7^0D^u(qW%ZCwt5r
z9~2Ql!8HBGwGTRy%+NNzpzgDE0sZ$$lvJpORHstPZe&K3U?Ck)0vw0DQ`^#Dd8T*m
zhwr~QHRw*|0(6JYO-I43N*n8(fge9u&^XZ3QKf`;lvMywdN9Y<c@LQRlMXiWL4y0*
zl~2*DtdC#!U7^8Rt>ekCtSTc8wmw>rJ<MHW5X4G&ko%Ew5!dh<-KG-}X8F~)IUFOm
zR4A~b=Eszu$*sPFR&sK#Ih3(7rc0&*Zx+9hOkTc(!Z}LaE(}O4k2T*?X&D!Z=(^49
z<a>*mTPw6DF|cZ^F^nWl33ux;hqZIO(uV>oE%~GJ<l7nKWm`?V-w(-k+|VeO+ZOxE
zz-s9_#m@Dqu5BjNi5Xv+F53d>>2D1F&P#ASEzJ3b4YK^@%f1J1eeD9IlC1=*w$Rf>
z;*ak`E}ll&Yt3m4pjn$U+4;}$tBxL5*BN&l)|vHA3fTfdV-^@02z~x>JBcoXHS{0F
z(bqr=OeDk1TeZUMLk32dj@c2#)B7`WeIlC*b4;QF9WrD}EH9ABcj;DOBlhdjT5`0<
z(me-wr|b*PkX(XH<<@i*#4sIvqG7dXM;JCR6*29_RDA0NmhjzoK)pR98++s{z(r;I
z)T>Xno|p84+zUD<rlu3UYh4-;ZMHreIE0qd3AzR90gVGRgZ*|p;-Ut{C<E;bWbZ4M
z|E~;v71$p16r6xg1z*qQF1^tykHKv72p+Bf5?fh>G(HtnRz^=0GTdIXl{FeyqiVM|
zzpV9|m4AnjxzmAwbx8JT-@qbzPM4^Sc^S*0eCso{s<lq+jyF+4K+HTN`{9pG>vU0D
zhuWD%II7h8r=eY{d&uH4pRb2J{AYt6+TW<d!{*G2;!;(F1;N0fS+Uyf8sT_inD^(i
zJK%F0J(Ga#*m*y5OT;U#R8%~PNM2ib?Rqa^hrQuewAaVL&2l%O`D{Z1Yc=*B-uKOf
z$mk{mABk+NWe<tGkx!wF1#!4-QkCP_+rFDlC*Cy1P0i~(t9@dog@K)^Y-S0tiEFfN
zOx+^;+y!>y3A{Zv5nZDD`&}Mpmo#dAMBp+#P2(5B+jM{0XY#1E(T;3Ug<OV#oM6<a
zK;$o*qtyJ(P4qx?env#{vT77s*B06NIXaPh{#^?}@?fY83P^3fxE&8PFq%aUgo4FV
z%NWIU6!0Z*Q~L_%t^-tp$-e}6$crgCuV#IIt@H8ZRia0RjCKlTeer%FT411{^eNTX
zA&wZDezQILj#>QBc*?%^5?_a>Z#84%>1Rbb$HLjM`@?F#5N<)_%tt>!a%N@aJ#Kb8
z{nbaJU!%iK94ox$I58fZgkQT(dm18n%<y_6hONX%IogHw@Cx%iu|QHDUO!?)8*?|L
zhXY{vS8<}-2Wl^Gp)04)g@3>jcy{l(AiaRuOf)T+%b2Gj^q5)Q*CorW(}qE?!RuYE
zp|!r6ZW!7B6$|L7E>Y<8Ez+rBr{UR-WHQOrmhV=judNg_qWmG?K-Q>^EtWwgx-jj1
z5RrN8P(zNvZ<eMr8bpZ9^+ErjepPF@#E@))-tHJ$iFZwix}Gz4Q&Ov{?4&}5#4hG2
zt?AoLsXH#X^}kzvV%_Fv<d;kNv(EgtOi*L2sk*r4sch%hUz%>qFnP*lB~3!aYpjm1
z@TXMYCr;T~{lR6=Aje($%47^&@V)vgCLL!q?sz%CTnt~r%W&@j4vQ+THR%Hmc~XM)
z#8?Ehj8ST9m+buyl&)H3i9C^Xa@MLjnl-c$=mQZ1oc!eAZPc_I$uhbNsHG1G6%pa)
ztk#1kXU&OnfoI`r$M-7e_&5Mo#)W-vI{6yO>h}4HhY$~WrgcS}5)bqxE4MNtEKfEj
zZn)@XgNs`bYNmwVudurei>QqmdWMzYQ48yzR4sJBd}dpcxfWXk5vrKNDEVQIS;-X#
zZS|<E16{qHnyJ7on~CQWQ`EEQYp)P}Ck}CF^rxu}aw*rFX<#2(WJeOcf=SF4++V=4
zl*SIP{!3tb#1~yom=1wu9;g0Rh=impOP!@dhK6X9m$veU6PgcC^goKEkXUvGKh)mP
z4%f*E7JKR9t!&bNV{OQi-fTC{X+KS(r#qMT=UuDdO*^8cAcDyLF48Y1pq`tjug7{n
zgkkk0B5+u<+#u?pP+%kvernF`5#Q4js64P&!(`X&pi6$Y05R}irYKp{<3-nuMFSNW
zXU>PrKh=4!FoM`p_qIn3@Es}<W1Efd;^5pCyd|h`M#?fGZH-zX-;H}ETL@7us<oPE
z$!C0G2g{Q;T(-#u*%O7`1!AnvRNa3e&^cLth(xy4!PmJXc#gAtST(BYCB-iMBR7@R
z=SQdvuDyv)ht2CS-f9-qh8@^d#5y<s`a9N(S<Cm=8f_K!R-PHAEWPg1zV?sLk!oTq
zYl^k=949`51@b{#lhZo2wN>X1QdzyR(3L>m&~zn|!AEyZ_S$Sd2Y7tHaCPLkbH4c&
zNJ=4%SF>B5Fp&}JSX#+oswK99zV@S!-5bXVB>35H3&N9@)9oIX-DiH^eyB!Lkg+NN
zO2>y}ssZC|;V8q73lzt;KSw&I?e|O1A0H-~C@9qKgv4F$KVhq<psZ*Qqn#~-nb3wO
zztV#9*SQBu1NCwzv>EAE#yPyLm0LuSPB5e(`rRA(L9|FC{D4l?PA`B>CYsAf*L6MD
zV+zUw&mYhw`7UlCVO&LYIIBrvy2z1U-e^4BaPc>rz>{vGdiA&c6nqqK5ISgDjg%;U
zt-O2BWWQYC`KXv{!60S~#U=q7sXECwC+v?)K1qOpMeQjCzw_n<x2+G(;$=vt=uibO
zXIS`=QYPLP>B?8)rc%yoOOl#K)&U`@kU9U|eX03^07lKlc{q3vtJj->$;z^K*?<OM
zKny$BV;KSA;kQIzWY~F(!}jH|OWXL%Ys4U9?QSVW7r6?x)IyzOuDjlahW3J`*9BBV
zqfVMPY2TUw<Zg=o5R=d2{=Cv`DeU>U49PQaPV|_M4fW}j04|jF+5_^z6nh2>`eJ8T
zMx+XPvL5StAHDJAYxF47SnX}<H}}v(m{SmHJ*SJv`uaYH7Nr^;$X*2_(+2#atkr;&
zC#rSw=!wy70>(L}dXy$0cObP^HG<AIl$#bmHF$f{XObcG_k9hS#GC}z=jJb#tj7;c
z%4-le4f>9vvgyBJ+*mPf_FMRzE;+6lDHtDG!*7VMwlk>Y_#&-TH$1Z^inp?2GlghL
z<05NJx0sC$m|*Q$*%zoq#)5|G!{P$bT(1+l*0=pA?#0>F6M_1q?0Q@K;Y`qI-uUcF
z<Ga<1H_+=RFft@Z9WPqG*M2C^pN+j}!y*hX)BbvQMUyrrpAWptxXPBT)Euunr*t-T
zu<0I4mp5cObEkaoHxfSRQ0at$5}(K}dh+0~=nPJkY8>@fO=zx^aDAOTK*}3NMJpN8
z+g(%$DA8|*7a&Ftp|&D&&$!Dm<{pwmu#wU!#LUk+D4T@RL&Ccy;@TEBEwV8`a~I78
z3j>0JSg#zHNoryneJ9c@6NVW{L7Hh%ieWLt0OQo(W~rOV;Mcc+4r!E-PqYdl6X3T@
zucdc&uv=?&-f?jvb%{{Hq44H#dPlGFW%o-_R5Q65llzdlo?Q%XK#7I6l^zXSZ6a&N
zbH=RcU`2sAVNjfiPziEJTNO{iG6%7$>w}*?3Hji<FuQBYWvBE-$9oFO7M}JtZYV;(
z^~|d=3MGSBfFlr+dJB_vbAyPp>U}DUM*S!rN5vM_RC0coZb0`AKBH-cr8W~9+sXxD
zojs3c=8s{YC5<nm#@p>A*7Dop-|rWRDm&Yyi49YhYqxf2N*4m@iUl7BIpM4n=DrlY
z?!>;Jg}i6OJ6MUTqtgpkfbP9&7+l}=0Zjx1JDv&|N<vD%xwZAv-l?YSOo&K~YS0Nv
zOs4}5#|9JeX>xT#lsCrBeF4)>mIEzGvd6vUThOuqzfN@Mr;VAnHIw#e>!U#XW^P}%
zQHHX~4v|M(<h!rFd%%enO*8IRqqe5;OvL{uqC@-n*UW$SBEzsbOi?&Aa(9vcPe(`=
z1i>wQ!Mk^eSaX!x`lpxnWzF0PFh?J0@VkXSKNs#cdESiS-~I4UH)fOLz{dv@Ml6&H
zWjk{aE<@?ooR-yvJLy+HS?;GV3~Zf)XpC=j2gN3^&X-*IHYT%O%fB+BMkx2$i*VOM
zsP3-vh07enQ>2ukje~PRmh2bnM5Cvf(t@|Sd@WlY!FEGl&U_;&mB)B<hc$)H>({kU
znl&E2$t?+Jc(C*kbj%U#4uF<JZ=c4^_~T$ZcxH|?$!g1ns9mB`-s>0af|as@|Ca9B
zcfDIJtwA<x2}k*Q5?Nry6V!bmui>B{F6Nc9Csr=5&DUSFGPj-Y-L8E0s)39rF?nU%
z`a_eYs<BC<#~2VpHfCxXOiXSZ6&F)eScs=me!1j`N4oX+YRH4dQPgJ)E8v3tTC0kU
zH9B3fl0V7BX9VMOCJ&$Cgdf)H*#>=%f}Tnns-HVP?TnzPhI&&H41#TK_Z+5XE`z>T
z>ES2{P5quOYdwlxGi(OYQD+9+(SUK7-EW@g-fIk}u)CX+?qW$vo%+OYmyw|8yN2bm
z$np`*Ho4Uq9WZF!%PRxMn^D}6Hn1q*MWoAe)|2by7-Nc+=NcgMk20%1Ecuimsp<Hm
zr2aP5Nan^1>^*$(6!(oJrUJ#R8Sp`$Gn@~vgSPb7)@G|+;Kb~VgcxRA0aJ7%ja5i3
zo?u~Kg^e<E3Nldo8cL$+Zf>NeM{rj3^W&kchDABe#>5>w;=)%DLTOmOdfyw0+rLy5
z#i=vOg5s~eRos%2TLjX-9MoP(5?es|#DwJRP3s%J=XN5HLIszS=OVpgu<*vHAG#VN
zeDN%{lWU^cN?9syl$=VuIiYQYA<$&~@V6pRq8Qi*wc>%JL2?o!z`|^JP}8yTO+Z4Q
z@tw(p?!LWzXQ{N*7R>G;{)=VM?Yxbsu#O%2o_ga3gT}UHzoxpp>Y>|QnEB1)Ln>J%
zQRzg5%}kkZ6)m!qW5p?}OD=ow-B_eK(;bemINST6Fks-*AHyV$D?<2YBA0Ksh?a(6
z0E;EX$3ijRLdPfZ<qWIW)n8A_!5sR^2!vTnt`+Shs`q~1VnU|RiVuUTls#A4UX)EB
zvYilO3y2BC1sfv9JZ!-B0i4|xKxF<&O-@?|i|>sSLi+y*DPT|oOTK3J^`Ya_)t5yh
z`)csohV0DE&>LY_$c|JdnoR;8<R?W!T83Q-^0if>pR!)jFETI~O4)yqU9BsH4f&3i
zzjdnUGt_a(+Z;@i49q8~IwL@Z@Loo5ekBs4H};}s+q^;ZlkM;J$yH`5o$Ns&zu}Tj
zFb}`qTE+~~TpJ?2GzwvV2+Pb_;l)#3YHLT1!1669EF|D-D~;n99UM}??CN_+k<z*F
zJI~Xy6-Khhvv|wAI+_nBv+E93UV>(4eq_K3aBFIyHnrqbaE1?$oVW3gaKT5emInt=
zP!vM&P9W@?!q_Ci3;`2ge4kw@5|Qnm8h1D*Te}uQBe}6xSP5u1p!p23cWK_DNb{b{
zHf?`(8c1{NlJC-*%g0g{4xsPxGKA1;FzpN`-kOX-`dYti#dXsxn$cN}YGMpPl*4g3
zSHq-AT})tit|pq^!_yfL_H*UfdAVWI>RT2s!OeEBAjxt;d7Y@uodfw!42rNk>0{T9
zEYvRy@8naX73jMHQB)l=EkKocvQgf+>h3}nI`HaH!)GU8T85T?^hmi@7j~nnzKOKy
zLg$XtednG3Arh~Im#zvyqEj9hDMCZipN&u1L%_omJ|*g0gwe#fmsLo;zGUYm;zU+-
zuDI`b57t)5JQoLQb))+D{n{H_DwVH-b>1ZNJt*?pmsQ=}+5A&%x262ekWQR4!2Yr5
z9T6e92972&Z_SW07`iiWw&$#~{RTo^v~&-z#xcp4>Ygk%DvC-ZgyoYqq`diAy@#IL
zgC4}?QE5?GFwl&QpuWj>;Fj5w#&||GEhxF))US+HO{sOvG<7ppTv9<ori)m^EJ@Xq
zP$4qW3pqFjskYb5tO~WfsLBdcl*x5ytn!-o8utF~=l1olU6abv<}{iqr;xUA8eY9;
za)5I>P_4oCVd4Tp!w^}x@$d~xDO#Uj@MSL#J^n3*x%^hEZki&C7&CRelc5aM6Hx!t
z)fd;oF&OIXZjhW%BB45@4LWRwp=s2f)f^(0MFs5b1}aitmFOC&uGlQ1j>%@mI(12I
zi%zIW3s}ZYbm!5)_6_!nskEqiI7c-N_PX8H4_GGN^Sxe_!W1ZJ5lBJ}jiS?xS^XD+
z9?Z{grrbdJ7c#%jA!~KB_>yZo&}?zc6oJARyHg;WiU;H95WeMdjnnou-t5%n$D*oE
z-JS()#KKD}t*DM~vuoP!?KYCyFPpS%po7$ngd|eYtpyBOE^?A)%7pq;L|Cj7^=}pX
zo`Kt{=fQ8Ta?HvRbpY`JurAs=0C%jZse$%Zq0t_HPXKbKc?Gjpv8X(=Rs9;(@||Js
zR(j`$^|^5!`{HZ{_|bNZXKlLeQ48O4PmDjt(~PE@S1KS~+6}19mj6;?-2^*veaT;Y
zCiy}a&EhMhTB}`Jn)w;-r8Gr=QilXjhXR*)kXeX1DHg^wKIE%V&<eRU1zRynKJ8tG
z9mxJ{J&rTRgBJsCvm=M0YUQ0PgF+*5yPdMRkoZ)He~%r5vNF68msyvc8jozI&khk8
zu(63oXtLbc4Ua^!TGBBg=?M(zI@~cHoDB9yf>n>d-1RDyL>je}Rx7-=h|dG;cXMio
zvAEzkbQPWfNizqBx6OFWPgi05^Wpb-M8RBh)RhHxp#URMG*AhYEMvP>T%nZB!(3Kr
zFr2XO%z4Hu+UKeTNnI6pYkx)n{0GTXW&pVG<F#XXEDnt4TjI~S)@Js$Os$D~Vh4zW
z&`_pQ?8<IWxlwKbY9sS@e#|KA=3?|An!bqGi2@&{y0sWj8b}R(ocxG+1pyy<On<|r
z4}Mr6ztoV?v8OJLYe1xVN?uZ8w(AW`%@bu*`QPprktKd-xITqtUbt{h%WL!dO)+ev
zc@FTP?<2;YXvEI&;Wu$O8rb75X`!%TrM0_~css4jf;QWT@#{zktvCx&hzWFdaMj0d
zr{<OPP<%Zk_z{0IHja3H`10w^c4wH%BoR?VVydPHIFQxRtL$ChvXL6P%VNCS-M|!M
zgo^4KVx+6odBku`(Ok(!IbIdBI=&lwOj(e>TVNpOxs=;Zq6RT&CxIr~&9<G7YI{bE
zCVPHHd#r4IUhV<lo$P9)xB1sdgizj29JG;Fj2k)Ia=xI6D5~SaqmPs&5t9Z{cP%BI
z<;g1HY=W4uRjQotF`{Q&T$@T;R^ErC_{zU`K~S)LNA~wph?(^IwanLY)7!vC?L+bj
zXx?g73{~T(idUBr`$Mh`Cyc`G<p*9egg@d=;4Ip7nf^d+**6u}`TBp|=F_$LU;Odd
zI`$MVkW11gNLpq8RcZ+)<4lfGLg^`8@SN>YT9<j3_b?X3t5}z4iwEPyRl!A=NiP9n
zhdt@W_wK0S8gmU6P$L#)<}@z(Jt&_1*q~ZVM^fu^Ybggt(BlXrU^2LhgAdCTCCJ}L
zR#z>_qQLb=G*z{n01pLJ7vr(LH|=m)<u8TY2e0i6PeZ5R+7#K4v9eMhT8Da!f_|rW
z=N`vt+gMZ?O8|crYBGCGD*l@aN*7)x6qdBPz{rQux1{|UBwjCzK+YU)pP4@k)_axe
zpW$5XLnbu-OIt;sOO1}-M6@Kpar8O!-9)gNf2+(p?|pQQdo3PbW{x#5dBL>Lt>x{g
zYeE5WE8*eNlvfq~dmCuqb9`vNZPG8IgVJiO@7^YSq;w<n)0g#F5n0t@J0lRK^_qrV
zA^1uVhqFA8>*7MR#?0>R{}TdHm52R17-SP80=A`EWO4fy)$nLaRIkN(cx92zJtR?n
zPF>S9{pr4sxI)Qq)!4Sl0DE$7y9Nv4G8rN8r#9MeE7jZf0n5U_#kt<QY&hfKgLzs{
zwc&H$ty&(CT{z|Bik`^ePD)YWhZW7Y{o+cj(Oh-%M<<|6_Q!x;)fwXh<>Uh)hrVa`
zo#Iqa7ut}KdU_l+g3gJPS(>%S<J_@gm1L)h*=vk~WVfB(e@FdjE`AI`63wZ7<g|}>
z+j9tCEupD3LR(Nw+5N=Mqgz@8B&|e>4KpV26eT@Nr$^Fic2}|&zVrLnnkY*i66Om_
zF@5L9OBmFvNLcVR-WIEe9$5fL8-zakMM$j~i^c+KWLQQ~P-#|5BoHqH7XArp{9-#U
z&Lcn2IO?$k3Mbq|BI--?jVE%BuYXV*`iZ5RBVALJ{DgoU%BQ9x+vGDci-y`O@)xE-
z7M1NU_gt-Y)my`kC)3xx0nX>;ljd{g{Ls<sjeB?TW+jSmmS5uJ<4=fu*h1LQc8;B6
zD!rXtmM0K7bKx08C$}mZpGr_voEI^CJn_bYUZTEwZ9meHJiFP4b85Oy?lQ7ESlwZr
zN|1MNhkSB&fKjr3d&c^18PL1GT`c*{6dfR({%i5u&QyyKI~y#P5PY^RLu4OnT2R#M
zLq<b`bi->oZa~1g&`*-4n0CwYz{I3hVKY&Cdl+h-$DNvjH(><5RyzE|w7<sbyl}BE
zEij79|E(aX(iYSy!(9m)mLKAZwbBz412TNuaawFc7Q5#|&8M%j*0zF9>%bITZ>((@
z+8-v)$}TLT)#({!d1AbT8J%({h3?q}QsY6{XzCZjxD<2pB^qIER#3)VH1l}jaa+JR
zRVdfvO!Hf%`9ei_y9Bc}hppF-1@vIEW7QUPey*>9=8ml43Xf!yeeMZ%m13FKtNyU5
zUh62wdL_~!_M}FnOUAc-)h=!7B~1&4#AFP;Axf9MXWnu3-10`&(#X@s#=Q<&AJZ8p
z=|V%BHqNE!(H$FMTLM2$!x-$M?v`FgOfD*>`&hNcSc&Nr80#YH!(a&qHUoVIt(r_M
z^Z99UU>E47Rg!)X@>K2`s0)p8ecN18l$Y2P>O%$lPrTZ_pQp*5SsT2T4vPbK$?S`;
z_w`p`IHe9cRo^IEl!l`BAcJiorTL4Au>D=V=2q;WrFa>J{)9;5aEQaBx~qbA#!mf?
z3k@3rhE5ZgO`SjX{@smx0_+p+Tzz%rb#~qH4<pxr5N<g3HQ`#1m63Uth{vVKWTvsX
z94t%9Q)q2LHqypL4-$t%@0B^N@Y`S@W|e@2sh*0iy$+d{FGm}3f2TY&N0#0ucSf@p
zduaGKYW%axASHcaAe~~Y&9HpI8v#sUn`jSni5@?M!{AlpK}cB_jg0ah6dVIP(386Q
zJ=-0S*X95s-^8gDif+$S;8lK&qzZNyEM6F8-#C9l&^x^mG&8ZT&Iwzi4K7w6^#r8e
z7ELc@3vnv-D=uyReL_7|gw8sv?2uire}yd<1S{PnPsYBWm9?r|hs7=~<PpF!Ksc9%
zVrZFV*Jpoilx&^1Np7d~-|`t*^i=66u$bxOr#)S@`OXJpAM;obJd(aT-y})K2|kaQ
z-wV(xX=`~f(QLwTdx{+g$GhPwnb?ApuI7)fRw`q7{F&*>e-C)0>(_<Y=}$B-N#lPl
z4Oi*MXhKlfASI(7W3(A1>G;yN`~r*E*ey_~!)m#D((rIo+Kj{hVOVPWte2L(OfB$m
zu=P%A`}}#%&K|ijqTdX1XW_dOK<X?0rNuMRU2;#tvy*%Q>(_!(xYWc=tVDv!2|FoM
z{(8-_I9b<y{mR2W9MmkL6vS=-m`DH*aA5<pZoj4!3Qc(HjuE)yg*#FcZ|D+{_T7XT
z{e1&MK9?6%$}7<RvG`~{bQwrlF-vn{TquUjQi6kqN|ub3*;d}=pJ_I7C0U-o((q@c
zhKTS=R2p5?#|kucOMZC|V$QHYrR6G<YJ`qpt-ZPsQQm7goH_Vd_#(P*6`Lfn?9K6+
z3dCNW7n{RyhFkY**?+<|_nO;I7zn2c*C)B?^Rrf4FSk2&=ode2iM4a-^E$pP)K~}d
z+l-6(Y)uKsUZik~YL7#-SNM+LOz+U&3C}=N@C<qwbCnJ)R!Q+9$_|ZINexgL9t%0b
z>ZvN7H#w9hvnIl8(u4dA4}Ar)OA2)ZLJBK$Cb8_JFv(bj0O+!uY1+!UlU-0A&x~dB
zwK;UBEljvwu+A!^POuR)%7A<NVOiJ<X|lr!YnBI7#AZI|<feijC8mY7`P+krsJ`S>
z>(e#XuX?B<^J_Hk!r$rdKK&gbE(z-}(mg@G<Vux!I{qh0{&NU{#exZ^b+LZV2|d2N
zB?qZn!VJ(dkTZUUVb7|(V0g~ZlSvp<tEobMc)a2wc9jW|t&(*U3unZH(3awow;=zd
zQcwZtid`}VsjB3Qy{{ji!$d4W{aA<!jOWmzb1oX*h$<Pe`2G)L=ddISuw~J*ZQHhO
z+qP}nwr$(KW#g7@+pg;SW*yOQ(1VQpfQ-n&+54<ThS%)VQ`&yia0T;OjbM{J8N<TY
zAzd=CYfg%~pv{ZQGA|~TzPew8q5ORj8`9Q#!sOUIY4t1B&^f$bS`LRI%XCvKZ533`
zF#!PTEdk6~C}7(i+EIqStg>J|B!^E#F7uYCrw1bac0kw@WY$s>cL9sMXdf)Cy{Y?Q
za<}kfxU<F(?s-_US8zRI4fb(aU^=3YWe~2HPU4;!(~g@SqYTkqITvYV$1PMgy4uy`
z4&}xRV@ic_>BR*$B35I$8ahHky06Wy7%)T)`E<+Kwpb;~|3Z&5-HGIHP0+DD%&FNg
zkDHQEiZ-}L<AK4Nv1^@r7o$SZ#-&-<HCAL}+r%;BWb%9o*{WufT0&xk14Ew_pO!B=
zp~|;wEMKS!Ip9g4iuz1Xzu=Cl2aB!WOcY43<(%qExm{Po5|it4M@#=l`gakrIZX7S
z_et<`s%HN9yud@}MzzsIp2zz+yZakLUmdMEoEm!3&bIqX7t%|yZR6($;^17goeTf(
znt(6RP5dO8gY#2ylX6pW>0vjI!35>X84Ldlq?xk`;>I5Ru^L`*cOy}cVymcm5E>2o
z&1@6xmn`&=dFYzAUym5CjcgA(5hzSG#~4oD4?0$kzRig)4E@QbQBYHh5)Q1BpR0Ts
z7IbXnEsu=ywwFdjy;m7P=6j|{245z*NLD=_2OaY&$_?urON=(^5?SSFl+Kz;7Jm@3
zVI1R8yKmM)4dqma)u(!;9Q1q=5J&JygGCEIzo^$XIlC6g20W?ptjKmO9yh}c;Z@tw
z>6k$b%O4A@OEz1-+cr5`qLl|ZONSPz<s1x>%6wyIlKGS7^OJNkrIM#IfDWT3>UN!G
zrT+2x<EMo9rWUL6+?BL-^1penY4|+fB2@|76V|0U(%?{Zk{GF|(R{?Dn*j7)e8V~`
z>tW8@8q3%qvf@oQBt$2+%KnPau^7K<MX>*PH62E#<I>N=N!Gg<rg1i{k2I2``P1{J
z#D|raEIx(ucHMA|qF9kl4sqV{OWul(RDn!h3o1EKbL-JuF)(H(e-(x#qP3F5or?Ka
zO?O|dW2SF|;ptEqJ;~enO=9MWHME9)`day2x3$EctSL|kgYeDAu_D!pSZf*)1Vcuk
zP~X$Y67I`*NSmfP+Dc8Ky{WoU6e7snY=@hL-i?X^#J*=abS*U&X?$zu#ji{yOFjq$
z=Y2~?4->#GGjns#De0;#N?6eD#3elRb_SaoKF2NdIM^Slid)X+-pMjC%tq_1vM=86
zma^mT%f<c;7UQbY7d+)iySl|c3UOqd?iHdMeeIWnnQ{{w#HZaR6SmpYlue4pD)16B
z3yGkna^Z^hAx%i{MMZowA4)+2uk;Q{`4wJF0aiR_HnrkWR``mfS}~rbu=z@NS+i&k
z<1QU9t98OZd}cB2kmHF6!YIeD=#gT?JUfWawSpt}OS=%{M`MiTE@1SZ4FCwBBxmP8
zbD>39BVp4e65#5hc73}QK()k7bQ@x(LJo06SDMeTm5gdya(t90z_)-_J#o&})&-1t
z1Vb$5oxiPy5+}H`ys^%~gWBQ9Si72SsvgTw-mxOiGLe(fov6W{WnaNg#X;I6Vl9%1
z?ycD8;|qh=C#iPs6K9Ujnvd`U%#oQcB`?mk2YkwpW2U5vTT)5wQ*HJ`2-uJ14+4)$
zd$wl5NtEu)2Y;R<SXvl^ca8S?Uo-1R?PeYb#I3F4h1VVB$5C8u9zdUL*bucHVjip_
zim^tDs1f9kTI#X#XzyJtA{M?qhDN<LGDh_VVBA~wk4sT7|NT*<SU!}8Nko6Qs5_Nw
z)gSM#gZYwmMFSEA<1E_7Y<j7`5<s6f$Lriwe_j+FS*$dCLkx^TqWRK60Y1etKy!N;
z#~WWfaYm?faE{bR){W}T@+Te=#>xft3%mSh348j;Dn*9JMN>uOkY6jPn9A1xH?kZ2
zXL}P^8ePO}t7euyHmIQ=5kYIsTX$4Lt|~%B3pkP)0uGxHSN>vSbxeOY@$Z)ntCl*r
zwOKfU@a0!v%|Cao*zOo2?jn4WaQ_JAcKGQXjx~#FQgd8I4<ytEkb!tUCg7hu<6@mt
z%Q-b#+<1OJg5H&q<0~S^)`z9UAME@1p^;XI$;067A3cOg_oP*&^~oa8JZBe&#NZ$^
z$a9h6K@{b>IZk_C`$23Ycv?XDsivZ)EER9W4vI-^(ze<xM?K{U%2Q6E%knT6+#7zq
zHC8!q&RFjwC13-fpvtAh^mNa+2{wqX8xTGyB)TcXx5;_cCEL!bSlP3z&Xn%wq#X&o
zx}iTE58*8m8lC)!&eH8~1k9W;T8JkLOVX>tg!tzYlig#Cub&1EldI9APefa-7Lwwh
z2Z^Pt*}>QDx^xQjaV!k(GTcE;X~wB&cq7$5*~t|d$EJEBf60Zfab)$eB58zw6sak>
zbWxD38Jk*Fn>SWtpo!QzW3avM0DSyo{`}|*=c6JJ<z}Txxpj8us(+^sl%S*ZZ>){d
z$7fstTNd^8zLSZp*r2XO*^`g>oG>ezOMYJ-P(+G9WEPR>Ev*%7P|GKxsWCJyyQopS
zkakp}X?z;x7L;tH-qu2|Uh`|JKLsUYA6=ye%>SAU5K+xUFnMRidxc8dd<k|S+Ue7*
z2cZZE54s;`JsBS_bUkpMAOMF<?_z`v2AazkW}pMXE^YvLgnoFSr#$c{mPK%yG0rGw
zyqt!NBQkGLCTD0>lJXrMAh7Pu&Ac+JQZTFq3G!>7De>N!=`|ck^<OV5Em^@MMNLFQ
zEs%VK&gmG5sjer)C#XZxAz|X};9l_Phx~dc{^idj?qzoHS6<3u4vw#@Ly8E_2^Qr3
z^n?!%D&gmZkZbPfqo0nm$(j)Gbh;rqh&M`3Ud}a65%6f2@I&IV#jwuu<W#TjkUbsT
z8?t69(daG!vF$!CHW&I5Cv<b46^}gUYvsg%oJFDy!Bc-7QK)qJ4=DGMtl@xt%m8ol
zt}?|CY41DQ7F1h#D}*~mptH>dBc8QG^)m0UXzTKV(27rJ37ZZ!K~^PwwMODQ)#KQa
zbjK)9W{5mWC8}W3*%l~xs$hsMz#n4H3DczI#Ua(tHwnzR3ED6q8J<6tUpK6RdDoW_
zAitmjItR<@EC+S5h$l!c%rW)%eX?Q|l4X)}r-=x=2}gWYhARn`h@=GF{B>11v|w-A
zI6i^mr5t)mtkKGs8{06tacZvV?Jf{`f~V(lQvt0zNd94kCMNTgZTFw`6=Hn1g&C_*
zt#feFy-@muB!Dh2qGEEHwrtF9&OQ`47(G3BnWiKCzK{L@2RwA(|E2%V`QQ5QEDZmv
z|IWt5#{S>x@9a!Wtp9I5mI|nfifwzDvJ?s51HoTb7k73E$w>g<m;o3@7Kp#ONKq00
zu75vbM2L%ol)D5>ki*#9@2%f&ul07fSuL~q?f2F9&G*eICyk637|(;719k)yPT1c5
zDX1hs)W*ieAOHl(I5ZF>LPG<F5aMhCza0>oaYQ#b3S9D=J%Fe?1V-QkWgsF&4jT@b
zg$Eb#R}esd1W11c2m}a3fWW~o(BL*H0BHmI65RX=D9av01UO(c(bn=R?!{h&0PbVG
zdcbxX@E$rk`teUH+<3RZR$-d~MggoChmei}`9@GKAaDmQO3>Z6IHVh;ILM%CQ2gBH
zW;XiO!5?UM_k{gDu%1E)@jbp_;DZ~`MqocUn6m&~f`0NtQUgF3+(NwmZ!ib%5cV}(
zpaPs(1#O6F;M9G{=HL$CiXlK?S6YEpRG?4Tl@HiGu%9j1fw`&Q+y9`k(Eg#ZNdJYQ
z`*d{%8T2S80BgdzgaLI?X!L{Ne_#U)-S`6y%;7nx@UK9_xPomJ0`zl(0Z~1D0b;Q3
z<~;D)ZWPo*!0W-6-cKWZ_@@f&G%?M}A>17U2N3W;_&=3{x`b^MS8t|Xy_@A2I@qUw
z;|<|M*Vlh>24@#j=ODmbA44jrzXJ<&K>R;(2p|M7NXV!tz<>zs0EWj_5^v2th4t0P
z>j{44*{y;E{M9?C4dStYkRV#a3w|KJcn9<l0QPSnA)kJ41O2r?zP<u*m(T#$1aJx%
zg8HkAw-7#e*{t4g6z~UdLJsiw0Y2V8U$aT0^xPaWns0aS^goZ8|1vGiGwpxy-uD|r
zK{?<L=qsS?AtHbko>7s&MHJ}){c4KB!+bA-p53arX7J+dfq&nixyt@ttX;wbuKh*?
z1HYLw;&Mc_A_RQfCus+S2)0##|NdrQ^eg?ahyB((_$PkxV=LXexc#7-{i6Qj8#X{M
zt-rzph-%4$NC9j(q`(;bLb3*aq1Cbt!(E&|<SV*@V32}%W&lS`^Lz>Q5ESnERfIz=
zfg6>Bi>`6~lug_BFUVJN3L7Huc7Q*fwu7dE0)30al+tUHFTD-{k4)_nHh@dnMCe<8
zSqdvSeVF4<GH7>f%Wp$ML<1POyEo7ST0K)pW89sJLoUG`JVvhr;M2pz36ug<WC8<v
zcX0oiV|@%30pu&s@8d7S7&!f!`Wj<w%e%R$;PYz)ULNYd($@+X@F8deVNQ0?o@fwf
zHM~tweQ4R?u@+WZ1J4`>2jUS1Rr83jJ6;H%q6Ax~H)GYWo_4vf*5I{TRCEyT)5-Wa
z>DQXlIHh(O9ja(LAFxPf@pKO3%e(NGb9YZFL8M(J(G3(VTI<D@jps{y0`9BH6>bNF
zmITn2J$0#KzsyBKRgArJ=ep%rN$r_0DnVx7`S!wJFj+M3AoCHLNqbx-Wpp8(CE^l3
zeb<$(d_@PL+(|8w?)~<4Hqm{$Zj@)~5n6$}6(tx?A^ki{32~r~J$;C%cZ%wD%iDcv
zS0qi}Y`mn?K9Ue$Im6+u!@JdBa_n;LwJEq6+@xJnlJ)%JVK3wtwPWbK&~6$r@g%X?
zkAweJ$fmIYh_*C&Qj$or(ZxP&?4vjQOke0=Eew8qah%&~ZmK-=ja(glJcQ<0Ff@dO
zCG4tNy`sw>M{>+xPJ=*GImzcEi=J?Z9*Dz?4!N?k?fK)JPm8VZ$j6eX$LP64o0~-M
z0%6A6-HU6FOrr4B(g6O>Syt2hSnHF;O*=M5x-c~rpr{UwEI9(OUCK_-Z$FUZ9?NG3
z`}1}5U4B(a77p43<g7~FsHiv(h5Lv2FGl*}PljTVYwwSZ^84z+1=?qZ;5j9KQU$(|
zF;hM)D{tTM9Dm%YDJjimP0*E+xm&dB&O0>wZ^=Y8Le0+*L08OzBGEiJa<OG&R(x2;
z`$G3Z+j7i7;F_^>5H0JemV6U{s_cwiPf-f|gd8Wf3cAKgO$o-Quw6|7ZYO*h`J}LF
z6~<U^RbyAO_=G+G@;GEzJqNtxX=tWmB~S6rtX~o@3X=Zkmt#MR^*91E|7K{6`e&?E
z52g%`p|JkgnHIndl3ST&qW#^cl5)$n<`%6YA$eZcuFKuNq2$3{<~{=8HFs!DNCzeo
zq|xNur1vb#kaDVTYviGp5C}+G+-#V6lDP*oiQmiImRdsJ(s6nylaUDMmq~rs;WQ`K
zuA-VHBj+KL%1oXK?cd2{jPHbMs8y`|oOc`2^;Iv*vuM(ONue4XL)@Z^k+5V{P7iMh
z&pqQ4FE5ebq9H*#1!IwM414X<*zySvs3@{qL<p?z(caE;)W@Ny;%ygNj5mk&e|Y|n
z%M(=YzQmcJ*av2+qsfg1N1E1~J8nR&Ns38vO87w)jTiYk&sHD{oAxgmD0^WEPQQk|
z&(Q;oTzNAh8OT!c$XL_w@d%Y9ux@Rx@zke;T*FCJEhA8vS}pOEtI3xE?8`U0g%oa-
zfWX?)%B1cK)Yi0~Eikq+Tb4P5+&B`%pWrzIAII%FIb7OwAw#n|;zh<d8ch4sztX1>
z+sO@sA9JzMIB<~iI|uPbQv}+O;GFp*7A;Jjxb-^;$Cs*Lr&|5IyK7RdP~D{!aek6d
zu5HE+91g))G+$YW(b=aJhK15i@8H+HgCryCH?z06pYMx#WNCIiSbxsoL}^A(rYvM+
z6wiE@Y1UK1(q8iC2S9K}2ur9}LyJSc7<vxem8o_Iq*hdwPz>Gq?iP$1UUAdrDhMoD
zdmR=*#8DuL%Y^G~jR&zQ^ug48c9ZwS+UtMx$exHvE^l%pkQ%j97c<mQUb-!=I~J?o
zUCHn)>GRLZU~Vq@uc)dvZ@>z3*)m=mD6%BQ@a@oXS;f4-4?~>$lbrs{dJpYKUhZKR
z8O%$kuzlmq&;m;vrCZe>WZJz?l=N}byUH<5J&<dzWEiZNPfa4{*Y+%FF^Ia_PRrz?
zReV~5@JU`~!;v(|;A3_;T_&Ko(f8iH@WN2=qLcfM=U!tVawJaqykvFGnaHDfy7lT^
zy1K9qw=MD9GQO|$61SAPT!fu|oDFil<H~9X@JqA00?>}c&t835$SJ118LB&$Q`p06
zIeQL)@48k%lb&jR=0FLu8{s@ztCBJYfnh92g$}-5m@eYNZ1Ub`Qzxp)daBTO33+>`
zd5K!wuN{XhNyZ|0{#}D?B^gb;vzQH&A?uuvUpT2kr1U7WD5oYk?gpXJ-^cPa7lcLd
zgYr7+=#izqFE4dI9%V1mEqtqk5yvPMbCwU1opo*C__F9|n<LZ6Pc#?dZ6I7Zw}`WA
z7fS1c_Cw6cP})?|-r+yznp-j)n%*?iz}<2G7fM9?0?9hQxj2C2!-s`;+v)xezvwou
zoyZu?_0?eaZxa-VY7}`FQ+pijN|vgZo6%ms9>1gMQ_jdu(Jd7Ck})n>RK%IWr%!+F
zP(=#Ls<~n04Z6uYZCf(^^-G)?hf_qC182{0SD#ZMohiB+L5HWdQh{DVb<SOVRmGFh
zQn$8`bOfoN(AAa6G`G6g@5jCCdd?eCEfluf5zCtnoL6Ol>hlxD*;ZanujqfH)%;NV
zEUmKMEx-1hfpA2;f`8VHEBqFd=S>~Is0YE6(*+_ESM9HVC=(XqRvMO0U=zB>Sy|r6
z%#yQ==PkJu<6VeNHou{0Q<1(;(K0_Wg@9WLp(M}`K}V;C8a0h?b-BjQDXmqU9I)8Q
zCP~LFR5fQ?o@ey^o>f=xbX*uDcH8uFps@$&menV6PxfFUpX>LYi;<l8N@zGFDqisr
zJCsl<F|UE5QTAub&_nUly|t6yg%4<a<@$6?cIa_?n-q~h@Dz5?PFEeyo0ZDkjifSP
z!<woKQK~W|)%u9E6l*3l?^LVNEIfVs6UPur;OJWj+uw=TbaKR=<%dxuqqH)#B?(rB
z@3@%3m|9oz^w2TN8f(>{qWt9aB*%>bYev@wXS>^jPwoAJN;5%I(cO8b1r8SXI_vbb
z^EpXF&A#1pS1_rU-hF`MwFwL-6}M^Mu#jcez<$cA(u|D=pvvp;2o6-*MScwgi_&jl
zjCIIZF2b~xQnLA%hpg$*Z_*T+Knfb}>wZkvl5!+_%7p<sq8TO{I$<)nZJgpK_qHpx
z$`DXCQb;UcMm8>8FLxt?!ryb#wgweN8V$l&yPY7BvQYDNRijpL_pFNqqP3fiBb6Rc
z3TF9W1;t>jo|Qr;-R9+Yg+Uy-?^=ojYkh%gbBs=5xhTmk70bmdz?l`7<ESZdL$O!B
zt1iZ~e>?jhVNREbScq!Tv^nb~@+sug*P~r5xV@<qEFzd2&S=xDX5mqz{*`AO?&2aC
zg(dcjh}Xw?<`dRe!Y&rBXHdwxYG$^?nd#u`PY|;bvC#MT`^OV*Wo5t7pb*~#H~czF
zDFi4a*hS{VdxvBjSv_WkJP8gIa|;}io{NQb%<=HGoa{@YIR-b9g(2D~6j7KLUQD63
zdd2!l`KZWZpYJ_hv1n#mTS1v0Z&{un`PZNn6(#k`5B0nY3X+WESemd`96hVJlmSDb
z!x@`^K^RI_aruixOU@C?!y1fsALelu8Em8>1q4)~9sA+T+fr~oQU?D3>a_NqmDSiV
zS-_SQ(iqC;AhU56SG_j?9Rjlz+n|QTU_m!smSlv04AIPua=A{IIUoI|=D`tRUmO+J
z<K=i`jQ*ClFVH#CT#6GE@)Re%l1h{3t^(?tUNfc+pPtLq*>(7;wGFh<)9QDRT&nZ^
zSB6`+>sIQiPN@>dj}Hd%H=qzIpfeSp-|)0nV|MA4pTjw*OyeywP^n@Ah+&%8RHI5a
z)3h#ak>ZrwZeV|rWj|Dnh+WoG7?cb&CrLSNGQ*oVTS@oPDg!xI49%3EEX#&>ic-z$
zloK`eyR}#s?9d^LfXd$N(p7SnM6(R~9|Oi&>~vF-vBHh4^lcgSceG$B4MNICWk}a<
zCgWsDB}n0`;QRf=+@c0=CX_d&bGW`AS3Q;Im9S3?-I!HfwCkB{`<Uto5EK4|W=h<k
zer38~X5!<=-zv71KUQXg@~bm{#?51?wU@CQ>Dzcs)!AA8(LF5!C|vW?=;>s6-;fN&
z++7`U*&P=&A8BSaNF9T~JMFiVt;l@S3yW4|#gRmAR=9twqz>>%or%(|9yRpX!E{rm
zC~=pristBrlx3S<vJ}>==H~Bg(Wa`;w?^fSM{&E{nC)vRxa|v*h548az`UhSZha#n
zVzM|$Wmw;QR{bsBd&iqKR=V+dn#pq2Qc<4pj(XNM8}jXv9{W^xGr~GYl&k|^EYr7*
zUF3p+<NtBhQmTGD_w<}c0OIPssIWiHR{Z=azokr93auBx=YL)~fQtDVJ63(xP$tka
zF9qxWYB4PW->S^0rfu9^+#$6;bIBQpr_(pdqq{BLKHB#19d^rbg%YveG3IyTw8?is
zZ@5R+9FR^r{96n*6)|@1g4M1dADLO0`ziIr^fk9KwepEd4l;Q#!y-xURl7J{iCA6p
z(kTQwwQqh}R1n$$YxSvRo;CjPn#Ht^>Y>9ljuB*L&QF|H#Lr8{!#Hpn7;EA~HYj#x
zw!G$GxfM8)S<`9cb_&AZ#;<&3(s{CkNcLF2r?|U>Gc_vVQWJ0T3=1TScY`Zh5*f_a
zqZsR-vv~ZfguPzJR^0$B9OEfBAFu@fi3(A}7LHtZD^Zea3?J;-1AN<O=UkAy1bc}0
z;G>gtZ?qJ?X~yC+m^#I&Yznbht&b17WTU<Z-DOCEc?E)D0c?4gi>u)#?#Y;ZIEfFt
zh_!MD)kv?pH5+KlC;iw?=Wo$m#L&#yGFJV5)43=8lST;Y)?_a-&_grP`P0sQ^hQ2x
zA=-;;7*uV3aZ6?vebkGOVC-Ic{ppY8u_jifOX*ZS6w3o-H0#A!#)sRfKh<gArN5J5
zLL?e2#ToyK_>qo&r6xXk&dpAev1`C8J(7#qC9jvK9(u{_cvn$<Z^+Ap<>KBw`j3Me
z<J6L!xznqSG8XgNba0~JHAiRDrOX!usQD{A>9bHHpHXh083g$Bp1i5_cf~;&)BX4~
z9|=w-;i$IR9mUF*QDRXLt4ht&;Q9o5wWgDj!iz_^^*(}y+&&JOrS$3Iy-ToB>r>)D
zJFUescf9yTnw$^tv!FhsX6uE`4}8^_lGSLh758MVtC#R={TT0Y@@4u@BA0aNA85?U
zw1mcRLKymGN4#?W4?Sf!eo{t1r8i9sI^%Nvh3_c^*ss<VmB`zvJ?`NMJYX=-%;7ZZ
z(6$+%Zm_V6^B*P2Kb7ves#U$8>c-mYV|UP`ADJu99=Gf_ANV<1dCZ1Jy)j0Y{5)=4
z2t3SmQbuiCw}yE0{y)4XfzFsaDTl&LtTTct0vV>Yjsy>Bhv2P2Z=)b=Z>T}OEP}ta
ziBXh*GcO)Pj;@+Ho`1JrZZoKxBAql}lPs||!wktut5Ir{_&66hnC$SABDok;Rh5FV
z>oDAi#=^_w9{yN2?!~jHo4%uEp!3P5zshn^%DyZ=wS>^JDc6_AfFSTb=y%P#(9RI+
zyWGa&GK+z9;$ZE#xoSid%o4K9*b-&>ws*j*A^gdY+xT@%qCOZ+oE6z9<}Vg&e<hs_
z8mYc}&)Z%Z8p}teqI$#Qr|c(9H=x><T-v2<X>XllBc4x*5M^OcVv{tC;f9YeMPLtM
zNVVFwUnZps$I(r!n@N*WY#a=wEc~{12>syrH#E1Mh9b7zB&@e=!Dk*?<ox6LunB(Z
zy?gBI`ar|Rr1iX}+R`~T11rSJ3TFhL&F1BXKIvYJCr*a?R+`6|cA(8WP3rBEoxO&0
z|6Io`lFi>_?!mQ|P6ZO}saEE@&kA<`!}v*Cy7l??>C<$*6PH)eaoy<99@|!d^D;Im
z&uk%7TMKbqc<9N5E&1`9iB5cLcsV_5kpH2ikN9#{c4=*uH)o34|CH4Ax$ucpqa-`*
zmAk34<s;+w$&;5<%CV0JFEoBV1x(hN5f-t0aBZ`Xdf)&|*`WRuIuR`W4``t`hmNxj
zO#QMDM<=c8OkHzvLF6^#Z@ISTYQNg8);;Oe-G+!?*K=A;CD-X*%*=A8LXCDBguE|q
zL%K~OI#pgY(o;1<-@q6Lp3G8<m#Dg{=&G+Fld;%}R$!E7>#<8WVwZl`ld!WF(a6Zl
zLSAgdCv~P<sDcVMgLe3ohD!gitn3r1d`Py5(~H3g6-p&umVI?>bbNWb{T-|ZSpqiU
zj#5YQ0Sm%wxzYEc5KDS83+berZ&4$7YICmS4oj{{DHQ!z?opZOOGd640Y<&pc{RU_
z785G|y?36U&h9$arg|-5C&eh8V20d?Ie)>9Ujqkot9g9TSrJ16JtMdGkUIYgKa-|U
zM3c22x$JKY-cq|gSq%-xxdQ6PJQ8t(F;T}hWyeeNh`1e%#BWl&Hyoho%$4fXR)!aC
z5?*%GScL0XkjpX+nNO6H{@a5m>_>dS8N0Q;(sad9Ve<f~$J9Epx+9}5w8!0pMzD#;
zH$8CG;jM`KGQUiHwfk&OCn5ULjP^GN1ye1_qWid-ynDi5<)r~;y&WEb&Pl~Af3L5g
zB>fN#fb}26gGF5u%VKEo7L8aDSi4}A=<VC(gEGz}n}ds&bsUghdM|YKd<Iai<qcu?
z20x82iz-9rS^FWzKP-%$Q?;l{6%EHm>b<@x%I?E85v!<Ui{Qpmq|~29pm8^xV(pLt
zQ6%XdK5$&E0N)k7=6+32gIsUo0!|U=^>`mp6W=_OrC?yTc({C$dxH<!bcg78bdTZn
z*WET6<Y`{I*P%>0&#p1Nu@aNY`Fc&TPLHN%Omf49$ILlW!%HgwTi-6=qHI@FRv;a3
zQL|FDLH@4WLSm{FB^qr#$3+-v^YSf_P6kM`i)}Vl>zo4@&2<=NW)_UeJg%4-YkS(Y
z`KN0#;Y!?S^y9$x4wJGbvG-W1O~Y(wde<7$lYG7CP1l!(@+VOxx6$l5?DB{Ss)A5-
zXxRyTex9S1Tm`SZ?R<tp9nO}R>Kc2g%NZWKD1|0vx(+gr1do?3wfLZa!+7Iy@cM*k
zP}#h|rS`*+_YOfLL!2?s{F`I0v-&67Is6zKn*`8;4QN((kRhQDj@|l3o8?8_Pd%xi
z*oWaIp~Y-)A|)qEclkdva`T%-DcsWkWj#aXOHIvGNS;C3qaXT5L)TLwl4g}f)aAlN
zRHXcrETLjK{oTcpjq`6Q5ENA-``x7lE#GqVq70DY>3Bo*Z9UF{nLuVW_bK6Q4;9sG
zkrJDTy@To9{-Cs%YXgmGFAE`5J*p36Lr(&tCoaT+Erm01GdYjEjwqg3fO>Wg9wwGR
zSc|8g@>!5kvTY(*I*YPX7(AsVX;dTz=r39#G-)YYa#{S!82Ic`t*t=zz<O(B#QO^q
zJR$VF4&;JR#m0ew&&6$-TXqyXcBs4WJS%jRZadZgv|Cz7h=J4W&$2JgIgJ+BecBO`
z>l?hM@#u9P$HEcth@lzshwv6u_&OsuT~e`{@qn43@m{@E+V?2|pPr!Lvx%9f!Pgk6
zEC77N+w;-L?9X4FZ(B#%CxhD!2fWr7X|Qdrj<*))4aCSgGv3!s_g3;==*B1G?s~H3
z6!;}<0fyUyqZYIxI&C1P??X6Lsr(R~Tdr{N!pi-$Uktil7+vN=-YG{S&h>Yi?t{^z
z*R?$gX@j2FVe~{$>KyE@Dd_}p>_gVd#b1=XUh2xk<GlU7dQupd`Yr8=)($vCcs-6_
zz>E(`lDLlqd{AM1vGqYn!4CNkNHu@<Tqa+mph7hOJFuZ#hEhjV$)3we&tz!0qy(4E
z?lFcj>Gq@6{?Z*^#AlMx<98U-_OQM;N!9e%Ga{BnFqX0;Vn+^pa*xQH`DdwV-z92l
z6~1mNyG9jle*G!)1VcMB(Ou*#+Eh~=4K)cZ;}a&-hpp!DwB~h2ae<ZhU|MFRYlF5J
z6SXC1>mC_NptAp9s`BeZA!@OZHhghv#z$QpgF^E1C<+d?+-P8H<ZWa3I2zK#2+Utr
zj<QeYK*)`W(dSg8NLY6h*Qk<570b~RNp{C;Q|hvQ(6oUG1l^<?pEPFQX=WdW<y|=+
zyvJ3{Xu2gSgGvVNV)FD_3jq?EHghoPJdyI5gm3UNm3&XeQ7N3>1x1@A%qEk6kqSb+
z#f%xEwS^@`Laj;-1iMB;-L<1mIuknirvZ(gV3mZgJm#qL)CwHX)lSsmU!ZBH!$wm+
z)UFBQRuRs-FgmV}4Q<Wh*_*V7B7-6dCC`5dAdZ6YD!u6mpiCrUeW(a_EEU0(1Mi+_
zqs)@S7H4k%xBX@PSx#x!B<^{oFYotMJzNMofLTf8RX`9w1{aU(_w>{}ZRg#bTN(%2
z+Ek{4nf|@B!!~!;C{2)tgBrR90v;Ztxl_k)&6w?<Ja5Mssl?`P+pNanF!)Po3_HjA
zT^c5s%UO52GE_Nx+=g854e3bm;)q^ph^|JX&B2AP`&H;th%?Rdhql*mqeIieEazy7
zSaAokyw49g|Ap)UJ5h(7(ab3>SdmI`im_>%_xg+&v&Sq+hi!c@@`C#6TnrTyh!tqO
z8}R_2NS7S@7gUE2r0xIam28avEw5x{{a<+{2MYt^f6FU5S(yL-KA)C<KA%f0Hkqgv
ziF6o77kA0vzq(073<CoUOY_^?1O$|%5hX5C>=e3DEiO?k_+77@=J}7m*1o%0tvjA|
z-Fe(~pMZa@ESx2@xikb&aP}bJ@Z{_i18S<OiVy;_v*VMqv(cd;mI8?k68c9p1b8Ew
zp_c&ZKHz~h%^@J}36g;c@vc;G_W(%MF#^Kj2Z$*qiAg4BVE2ws(B9JvgPSP?;Eupr
zK`NYpQgHSV&}j;4f{V+Umo~ue9cKFD0L&Pm1H?r{BEI9f1vc;wU_gWS08|kM!Hj+#
zh%rEbEw}>;5kmgM>OyN5B)}2v?AY4g<`jIP%YmC+L87?=aHkjqD*$y2<meWp0qlnc
zs|dUz^j95?1O+);6VUxhvgXzx+!CBc7zhRc0~Ow*gU73Lkd9#9+jk0RD}oZ(Mg;nu
z(m19Eb{p931-x^3_~PE$-RTDw-1pBBG-wF!C5E_x4Alm%DPVvQ&=m|t-VZbY57<lV
zhZST%4+r;8z=nVV*5twXn!o{3NLT?0z=Qhg=0R^@UP9a-JO>K>rc1nZ&j_cR8lpAD
zy}BlJ2!UR?{W>a`MYt2*<n8>)s=?j*gM0JM-ViiMTkD5zcz!Wn3LVbXDTH$T!{&)#
z_$_V*D+mD65gPgy;t7aA4loRCGxW^e6IV`r6rXh3{#FR2Z!Zom9<VHoh@fAAdbAHe
zwF!Cz21t;jOW3#btMbJ@c(4zsA@m0#0CX4&YV1S)37m2Gg8qxa$Gif3AmTI7n+M=#
zx8FZs2Jg)D93b5DXY6avr!$3R$t6|4@mux2-<6384iCWo-vO@&U=Iivj{qlla=81x
zrNAOM+tqaV9aBZLI|kU>|DoOtXZxl$ZtsU2_<D6}0Qd*$${GT;CImS7P1wrN4#57p
z+kf{j`@sJP>I(Z0sEhdhm#kzDD)?KG`FG*he~G}GV*1fO82oP+5gwYr^1dPL-mi5P
z{JXWKDq)&}`seR56_EgZGy!b89|1EsJU==6uf9`7G|&>TQ6->gfbCDww7%2!ZZ%l&
zAR!hU-TbcoR4_XG>+c!f3$)GaonHsf-2*-9__5(Sg8%3nwlLnxdH-WVcrY%F|1EBP
z00F?!70ko17sfX*5`f1e4^0K>@F9FNpiYhh_;4`vUZI(OKx#1a$}J-_6mZ9pU-S<_
zA$ZA)J;_%9%ug(c2!IZ}wH^FN`;q&PPTFrk4**@YzW^UV`ZIq4KmhayJO#vF(>FK@
zki+^nFrN<WC;X8ogV<N#uPV>l^7?xrw?D_{e5U_CKRp7-0U)69%p6C+h>(r7kQRMa
z3FS5?+vxBD{31P!2j?i1EIzT<G5NGwQcH6)TFNf<^y?i(#)ezP1c&aCPd4wf-mir6
z<%IvNVDa5h)+mwT$q$Yv)%qto>NboY@n7xlRi-(SUCvh`O9gTM_Yg`;%k65|;a9_|
z^E_!ca}p{=1SKe2RqJ0B)ud7(A=0Au_Nq{eCXD%Y1P)^3N!`|Xc)e)GI2mPc;c}WC
zzsS@hz4s&U-}wO~L3melmx3hi`B6k+j35C$n*gae{9+}0O8i^0*WAsU?D8EndI(d!
zCJfmIl5`Rqlb4IzqW5rhxOMwcyDqc{SJ1VazW#}%A1ozl+KAf>(>l(-?T&YiA1n@%
z=P@UK5iqQGiq<}$yNN2Z&Ao%k>9t!bsSSBdJOMXU%g)!-9PwxC*wdO=^twA78}QA-
zsT$bCvP<9NqTV1u(%~_mU@d8sVz=K#-L=X6YO*?LqNSR*eO-%<B&jverlip+osHQv
zm;h2BK8d(s;c*Tk5~tNwv#jV){7liLRtHFk3j%X1Rt0~;aJ5lYoO3j@K=uv&I&|sM
zSFcNZt{FdtF3g`($-L8_CtuJN-%W|^^h)wQnVWTsqL<)L#_W+`^jCf<YNIV)cj!{z
zj+j{0l`^Ow?l<f2S!Byl4AcuPVO~5ebxD-p7hzS(ad3^MDO2;AA_wNFVj#{M`C{H0
zWMg{k6-0Z)hhPMlo^kjuG)>peOySxgShmXBpTx@p5n{vC3Mm#%;P>$w-<;G<{ywha
zdO)F#c~IlOIpdpGJ$qB@5eEW#iIE6{YmIVzm4q>(MoBN&EQ|1UOaYkTWm?G6iiN|+
z68@y&?_isQtkQZ<v37X?CO|(p_6I&PI{H;<wpJeJw=kiDp~6J-?1=OlHqO5mId^)x
zQYmo~qzt@CBNyjo1R);VId8>Y9E45LWTKCJW<}W(7P$DoSqqzB0Fo=WdnN@tyiwTQ
z`Hl55M!(%pE>h&b`j9(W6n#~Dt*UO{z(mWfV|CK)h1FeOU){NP3cM#VSyQ>@1JmoB
z)-Po-UnVnc?@?bqMV@j>z8)!AkIEm`$O66^t2}&_OVEpqvw7~)5YR75$<yVm3Zdu|
zL!J|LL$4{oHlJrz(<<wv<B>=3J3r)U-*&@rc%q|s%sq1jT~IP=2|*}UJv?96ImDmN
z&*E$g<^R*f%Ct@YEP*aOv?2WZ5EDt3U<c9U3wkr&o2u4@Vp=O4>=4!k%c7};RrGC2
zT<;-me}`lFplv?2BWKlaHnNv0-?-ZE{1wU~Djwnc_n;OCkbO0ZBN46s_mg{iB8#vi
zf6witB~~g@{0XSVcSGoQi89HlpljTaOFgDQk2uB)wZyv820qtPCOCZbb~kcNS+aa$
zs~&EET2~xg?vQQ4`=HFZ^S+>lY4q952QYoAB-#b&_aLB+2ds5(Vaiwg&&eQndwH|3
zYsXiONlV6=jy<0Lj*{#@unDzpI^{G@Zch?=^xiCsGsnkxJ%4pgM5DubO;8J*)ik7g
z!X<k6Ygl6xzPSJWA+|T$bd)ot9m>eZWgmX3GX$@c)wQM^v2>_yJkLytz(bYj9kUyI
zQoGqcOlX&XvZUDbSng^_$GVoh0o<wn6V@}!O&kZckxx~-vtRHvnQt`et*%LA@CTo2
zbn|M;K}J++MKmCkO8QDf$cD=@jwQBE{nBu>5hbK8=fq;GTfqe%+a2}eeO$`HV51HZ
z)?cZQOU>WZ5dBz}c`jpY2lERl&oRboUf%#)l0XwQP4lZEDgCGU9U=89LVt!guEtdq
z;?LZEVo}}IkmN$?6B8^)7U8(%C-oC4%Juj4f;MEvGvMKNkGJZ*p~zW^d`3VGAE~sT
zjFy7p-m>KIS&czh_ARC&BY(Y22ne2({lxHi<s$fA<8GFRXXhA2|9e7<@avkkhN*bf
zO3p^{x{cBUzUwx<{kjBKAkN5yp(X99*E;g7$3m(7vnK7rUTH+l`9rEK6cgvBLf-9C
zMW$`{th-4x>D+$D@-gP!Tn<QS)lBVpyfQWBK}rj=6!MbIgSs7+<0sAO=WansE<DC~
zleiNj!sy~W&U#|&eoQn4M9eb)WhpzKFZ>7E&OZH9o#)-Bif}9OPT10BYLmk-FWaWd
zR(HoZQ-7Rt5391@7Mk1NC&n&>4;Kzw-OafoLtOF^H%+h=D?^?_?NjKieX=MpB*7+y
zUZF}c{bg)#yL;UWT-N|5li+jQmM!*=1fj?*dExI?uQ-@|U`MyzxW3Ru-fV!pF?AGq
z-6T(RNXsHqS2>jEk5LIsEos$*8#dEp&87C1RxRfA(RX={Sy4AhrjZI^>BLXg#=Gt8
zAR?83w$uY^wF=}H2&k7QO~xwXOoG@`;SqYVvbWOCengQ6%D%~YQjy%gge~tz6UmFP
zu7R(}Nux<lo`RFA6_zvxsiPu9q>`>NBtidI-o7u()sLkKIPJ{dsj{=r83dFY_O7my
z3fu2Ol^j_~MZW02Wit**3MPPWW5fG+=?UotrbB%oH1IdJjxCPIgOT$)yke##9QzHL
zL-V$(=R#Xv)RqXY>O+s;c-~m8p4Uz=3u3-5_cxRa6h4aK)g*c1gNn>m=bozyWSRLs
zwm}hT?F-2#yV!PD8l<wVw0Z1BB#B4$Xx{_{G*+F>#^#8#{!+~A{49aZs+V?Tn*nll
zX!Kl-mS^aYEg%Gp6FG<Z{E*sf<3vaA7&IOpT8@NFMXIcnnXT;CCr44TN5tC<n@7uA
z^=*{uq9g(`mQpHCzI;_{C?O9-w=*k1CJ9%L?$@sdkE^N8=Z7ZizVn|I1PeqL&26Ie
zf=d*EaVawyf5(b%`Pn1#i`Mo}X07(&aUyTpNqzw+*ZJ%raHsN;NgIBJEHgPDln(Dz
zdU-iIG;UXGPc|GAFi_rkm%@6--XV#qT^%)b#4}92pIkAWjyt18b~Kz$JtOfmsJwjB
zKbLoKe4|Cho>GllnLAnaBd(cSTg+twMGkO{>Rp@A)x47^gPl40Fi?o?zRgiAjD<E6
z(Hck?!yjWkEI)kKIanGj>-_C*D`}!AZ1^sPbzA>_V!c*3<o4U&h2R_P?ItRWJlsDr
zVu!JUZ`|>$-}$gwmwZRjJ15cU4HmyR4&{mHE=CGM(zMBoO3k6rBu|V6UgH-}R*1KW
z<ime8q^Za1iC4MxuF5zqL6bO|pfDjhYLU#lDDZUjRxWzSqSa4Y16@S+gSapeFQn|&
zfNXuJXwNv;QVnzW_xt1o)t8>SC0ls8MY*X2M*8p_EGkbX9&nuA8~wf@w8HSSVGrpi
zK^jwUbPZX@fv71b;rKbW4CS>8a=s+?HNMFvyY$uwd5&5Of95DCU-4FA$PFHFre~gn
z2t!IyYav&4YLW)^L`trc)7w?+Ok}l}vq~P2@t7WJQ$OjRNUk_^SF2$XTLYoGQb3}J
z0jIod5)PHW@f%@sS`ol_c^<k6{KcBiBTf2VwPW~jCi`<Jb9EL!h(58<L7%=^8hjGM
z1T9ll(2NM;gD^A(jW<MmiH)d}ccexpuEC9jJ}y^=eIiL7sLrY@0g~+}gWv3{I3xdl
zGtbp7UzWkX>N#nKX8G0C)?3zMQ@TyBLoG1)1kqYZRo)(crhm5sips<z_>+bulTuNe
z_pF)A;V(|r*hAb^7wWW~9I%mXu3y-019Ir<&LEvq8RTRTP>*yvVorHM=tuu`44U_n
z!pg}9hQjUiB5+e|bTB*3B|0C&u(8SXP%$M%RW-~Y9$TL^>)NC{5044#>_#TX<i`>t
z`bPv2wVji46c&r?W3*3Z2FSygTOQyWx~Wrd{T|_7_@u0UIDfuU=js94p99BuDQZ4U
zXIc`;GsH(yJGEZA(p$-HDR8SdE5&)Ig5=LzE}atD8Ah@U_LmvkMa<WFLVbUL&R$q=
z&dsG-cBQSUKb|UDcVBVm7!B5Gd?1bv0}%c&otN<kZpU@j62GS?_}KKxVv{WEgJaFM
zq?U*Hr4o^=?^}V5E-Q?~3oeMLaf|VR9k)!q^yW|`e~lA!j6(ymAx_pI_pgKs;Q+V!
zJpGjq^F_ZcQOC#-m3-6#Rx82Q038UqW8>~aB4>E?kCtuGJ6qPh2#fOW_RF0g_H~ue
z^;0@bwUuG!C}#Ehd|s*hG>rs4HS-yZaX&C^pG4v_S0ibr6$^xvzn*)k=kDPrF0F|0
z6DTuQm+jr4v>K^CLT=~EV2Y+4E|7gg`#b1gOjWh}LtXo2Q$sAy6)f79XjcO9bzRWF
zmFad&|K@ui*m}Jg#O_#AmU40BDjF{cLxJvdw24{zEiB^<Cm%Z+DNh8Uh#eDO_TO)1
z_*NB5UQSqTkq)z57ge6X*`7O;Gm3+mHRG*UCX)wd*+OiyJt&vj4)V+%J1hB51E^gl
zSIZ2nRbF4V@S{YiS;*6O3j2y`SX7*c>24Vu-G(9uV&%YE^I&YfXs3?rpz;0CWPGs2
z*S_)|p?8xj-Nhd2vHc1@3cS}+PL`)?`%!b#s!u>!G2NWO=Rm+DaUBlXMWejXO3-#*
zeF79cD5~?kft2Q2jMO3NOQ_b+t2Cm&)*M`zj9QM_OzWKAa^CC5>oJ<)6ow!>(2G`R
zPpY;-<%7Vr!u4m;)}^)@Rqn%-R59if#S`qY&=|>Rde+^n`MZFWsf#S`MQWJ<*`Pnm
z80{!uM2#@Ei;ctd_*1Y>*j^hA4Dhy*xY&`sbdvT}A5P4o4aXC7zu#{UpDjDCx2J%l
zzY;wQ`n}~C>X#Qd(l!cWYsACM6hwK3b!Jy|D9K4zv=b=aA-5-vv2U-`#rLjY@)ti_
zp-OEQXeT5t@Z@&K`Imv4yzSEZW}67diJlSDM<EAlUF13u2z1K>M-Q3hLpQY0{Pc|k
z!An9mX%;FUEPn6&v>BjdpTs1mf6R>A-qkh8qHVnv+2%MOf-~%=KV)~u<|DGa<E7Ac
zr(DkTjHitl`fSDgS{hy4_*}y|=^vWLB-3Ayl8mt<$zgSH=B+$Q)Z=v}=(_JEFNFqV
zJ(Yi;s$OGXsvIwyD%-TCDN*^)F@FwaMFwNJ6eM=3-&t&TSL(Yi6;?OqD_Kp9{i*rn
zx|1(luOAYBC=xrCpU_l#T}||h#sxU#juX=B)P+it2${0Ao-P$T*Rp<#VFEmZaMOAK
z>V#CunNTxhXlarM#OWMfq`rEsZMz?UjtSrKwjZ%g3=P@j`2KFDnJsbx&FcQl^-Ih&
zkFzVA^d^1viPZ5GMPzloHH@!@Ud%8^RBT?k>m~tiXPg|B<HK@a;a8jc$QkAbaMme5
zS|{TAUQ(P)JucKGHL=T^HcQGps@po&v-eZiXszO~dtt-r4qgy2+v(g)#q}EQ@W+q3
z&5UXLXfrLozh79Oa9MY~XOz#c{VAVFiE=SEP3i=l6BJjHHf&Av6@F0Z%WT9-%&O6<
zfoWFWgJ;i$e(V6zjoNqSz=7n=1pF+=(ic0MM|8<OXzr|(lFh5s_RT4)I@d@)iB2_F
zR=VF^D?kgyr_9kRs=eYtq(b#ZqXS!hv+oq1fzn69hi@3+l_nUPps*ac1xKnD@qg^}
zIstK}pM>p3C+2iKHUN+Fn-wJwRQzkn)d148n&)5ieOGu^06=Mtm<(|Ga-7|<U#2X(
z5cCJfiYNzviy0i&Y1czdJyq13aV>9`A%8~nDg@)pD@a>Sd#0Uj%n$~(AQ)BsIlwI1
zs6a-qN;fm<GoJq1E4EWtCI3TVXuJ3{y!?oil>Q>5_mEk$Kg>-Wb8p229l5mBM(GZu
zk)GN$H$|*M-*VZHBi5F}>U%&}=1%h|1S<YY1f`xA-axP8dmK$O-u~LQxA2>>m)ci3
zCF)%%nCpT2{1wR@L6PP{(X!jSmGGbW<wL1eF^RazjJu(0Q4~g3C4u3XTMTU^G?%2Z
z1m2TTRO4lOPCDl8Js0aMSC7{jcVlmb;ZQ2azn!{M;52Ki1z&4dje-nS8M*=|J8yG3
z)(cnCoN&Ab&HdK8atwmET&EDA2?%sD0&0zuz3r=5I_x<Kr)F08PmWP&ZazF#aYLm*
zy3Rss+(00$zlP+_HxXI&KTYXH#7wCH*XM^UqbW(E!EHN{ETncdGi{JNKcx#7yh_`P
z_b%<;GLoDpzuTYUb5T!o?D*lWPED;e8Lc_!Exw~0*aH%I7WcQaTC|W+CpkDyr&h9>
zUqPgNnxPgcFH=G6L82cnk{mUw096nNRgSs5u+TX|2hupjbKHaDFG|;`7@Ckw#{n?T
z0pApGU7ryxuk(U_9|y&b5gB=kci;(x5ZCaZtGD#gMB`4-u_=@041xIM{f!rHqWyvh
zVz&>O!(?<+VrL#8jeTtCe~O5^Hmaa5O^;LkrviKS1FRv>qabL%*r;q;FGZ06Y#p`3
zA?4kz#(7hJ$~rV;;vEihnV0=;?|Q^ROKgSi_vo?DJuyab0SCNO3Hj6*__P2knyO;V
zsva^1*P1{T6hDs^(nCf?>f?!7{xFfb$&D(*@#uFyR_<aW0)wk9=VdQB@uI7%p#8y$
zuV>|86MN1)Taf$25~t8j+5)e&PdfiL&%qaMwb5-hM~-a*{&{V;RqW%XuGd~LO=K6b
ztIAnzRW4>nl-YEvE~s<oiT$xdL|$IRO6e^tOe!0lhO%<&tRe4Xxoa?yW)0Tx%o}rk
zG1=#TSTMV|9qY-=1XH(9Nytye^7-xFtd8XwGPE<FVwrFrLV|K<_b5<VY4=ixGT30s
zhE2Rk^$+X>OD)gwy4EtR2%~&6KypLC%3xZ#q1Z=v{g*xad=g{Dn-+Gr0J&jWZZDa!
z*BlIn@%wCS+woYh!@rB6O1^mdgo!MpfUuAL)8I8I*IoxXRl85%XNW~HxgeS{Dfd@%
z#aR(=XgcA=_@Bn%_j$=OsibET$Z6;(t#Por2Ru`v!QPTpJ)`p}5JR4*IQNp;uHLP{
zE)I^AThhzY=IMsC)l8`Iteoz$>P&e#=)X2NVj}si$?=3jNi@g>R-0xBZxjCE_C-F0
zR(n3^bn-h__I1sdN=7Q`@*G_WpP3hqv0BFuE?C2EsmeEFcgm~o9cw)oS%k&}4_nOI
z4Q^eezLM9a!HE-mx~B_u_DZ@2?nGT@c4vf_e2|K{<m@+g;kSYxH$TPQN}R|q;Bzaj
zx0YoOMW0nKU>iaz-x3mCF{mmb*nqc8imxD-HY@a9H9^Qk^qUkVSKj<xLNQfc_e7pw
zoUl~2>_jDT4{d%aYnFXE3dBp++lxIH5rF3~nh?bH6>MqL?_5gUAFuqD>!!xVR79b6
z*7fcdn|hLw(<T`Htz=xNS~AK}9b73HgucXrkw`C4uU;4==?Q)u>bmTx<j1VxY3D2>
zpPn~OTqmw$Dp?WjJM`Uhdd_i=g&!7DfH=OI8}*;e%~L%cKqE|!!#7&&rtT%VF{`Op
zsRDtxdwARMl?184rHvOLpSC*RFt;lEQ*vq3Qee80LiNR(*_y#>I=2(?5|LzBIG=8^
zGwVqpoSK{);GcnqaV~&`*FN<4h6dKv`)rn0T_<YWVcG`P|HIfj1n0s<-MX=DTRXOs
zH@2M}+qP}n&W>%{wr$%yU)@{(t=sq;r?*zsYOUs&&m0dIar=<(iKQWgb{RtILecgk
z*1_ew(qyfTQ@840u2Vp_8<s88Q+lTCls`o;F^=R7)#@5!r=(z<S=2Z#<*2Q-%*_>E
z+7+kl#(!1RL;Cs*`N%eZjyo{~@45cszId@76SSxdv@1Q6c#nZolXoT^#7xm3H90Gz
zkcx%=?NsUKVf`*tRab0_=OZUtC$Z7pI88B6C5U}l-$5l*)cG~cFWO!gaf?_vc>J$D
zT`>;mX>nRK&~7c{wYreSI54M-PASWV411U_d%pZbJmLL8HzBM<PZOlvp27EJPpYl*
zvYwE;Vh071P)`b_A3El`i>iHx>U}Fdyn1a7>0UFegHpd#qjnR@()S~^xW%b{)6?v%
z#-)m_H<i)m`0GEr6jOeD>N=a5?IgA3KSmvrkw~CQ<Fvi|YRZaN&E`=!P&FOxJLSJ#
zFhd3ywdY$taD*pVKUTFx@rtt4Y-qJa+^)C^T9<e7sw505&MKkOW)~S-=lKf%3@y!+
zy@(m~X2d%@zI}O(^>Jz9)R;{Gc5%4l^W}<CW<J{Iq^SfrJnmH+JCt~(dI#CVlmQ8u
zNfq;RYjo8Y$8{iTGT;o%h0@XYv$TDx(Z>J5o7S-0LzPq}3xOP{<C$F$9`T7iho>J&
zwKpCoB7x9+aMN@tdx|AyABp9SvdANKmzks;SL0CckBcv}OQQ@YI6i70GTC>}&Q1D;
z9)eqN91J8~bMn)>0<0y$+eshjV<m0Ae}AyPzf{0XR3guzbvu(vi?C$I9Q{1o>c#Z?
z`*3!NKA@I;2=Pm>RHxtx`Zw<+tw`&(<$gr^{(eDI&s<9<`?mBybno!J2}kp#QGBBu
z*WTCJH9dd+yDicRF^rbB%ibQ*ay8iLIg!C@Fi3!1>qY=$5MKFL>cMi_<WE#m+E5Jw
zU!~`rLH-_%K=M)ZJhG1JpdqH<I1h88ow5ZY?lm)`wdC`Pcfgo>8vL{cZZ-!)>>*?D
zQK#{dT+}h?lH|~>)azl<iJ#2YTRSZ>pTOASQ3AL4a^c>07>n7_$L;X0wTP@8AoZWY
zT7fh|{~ilH`rl4s!gF6>XwOXQeoy%&>63ve`gRGB#Z^J&3n%83V00WmzE#TYTu9A8
zaO_~5ucg78W34B-=G=cm)D|NagEYD2De%XvtYunovTS$4io7xSeV{_&@O$IdkajVA
zaL40iX@G8*k5nh&tXYzdx6w@cpBw^*7MuDXxxjqQ#*&>9Pf!++QwS&@lD>DCvnWeI
zU2hT&vH=?Blgdtf`;4;a>{OBpwS9?e5U#z73Tio$Sv1*YTfQD<J++HGg8|r}ARhXt
zLl1XoT9G)FV{CEi#<z8YRcuFPGy1`=;E0(v3DVazbfQSCq^hY;G_SIZ{<SQ}I#8*C
z>QBXhQ353kZqmkG;@Jycu+`~-gXWLym1`8u>PL6OgN85z@r(D80OzV(p8&@#n9Pc;
zO(U%gyUOO73Uz&2oKXqB2nA9|_~%bXd*DA!f7V*WoA9vqVKULq0~-$oUx-P7s%$_%
z*|}r#S8a;3&#j6hj{E6<CH0#P=0;=<H_W|l-%7j~Zt9mss`>oddQ2C5SP9y5oO19D
z-%`#$PCw$kPUv@M;luguml&`i+`!iROP!avqNLk`R~j{pg8cPC>A;e~${f!<Eof{V
zHm^#y+@5KZAyN@4K;1$-1hY3b-6e*_4|;7X0~~1O@9i*B7?7+r4lPV9Cr4tXs=8XS
zOr6u1&4!1#QtY^6iHO>@E6w*1dWkFIV_|2i=oD#Y_INCWJUq$JG0d}ReE*836mKlx
z2WGu2Tzm;3IYsO5k#Q7zbim!X5JDROY7a+}R~Y{aPFe2k$H;4l4r0gQbdKE*YVO--
z&c(&l>LYiRs1QQ<@)}&UiExP{ZtukUt#X30Qtd?Bn5q$i<3_&z(?lQ`L~NTb@lrT(
z6W<;ZH}DnF<UNQy15N>X^GV3f*I_bPFu5Yz8r1uzi=!`F{GLG~L$s7v20c9nX!4T<
zN7Vl)PUT`zZNoEvMD1=!J3v}!0hrK1aNyuZO13|rt@fZjPPI3bT-?)_)eMF^DEk12
z+GX5*MJ_`8Zebj}6?yA;qy;daw&SClr!qad%_275KdV3y)eMEw5ZAVA+sjZu`-^(*
zC)ZBJ#lc~2J4mz_##y;_*{m<yjSt5lRtT-8ueMcp$y&keW9RpG@6xb@IrOo=+U&kR
zBO=ZV_!5#rV3)YRp};#DAE7N$d-#0!LhimeWAyFxBHTV+6m<P9f};?l<1Zz<H#BV@
zHJIf6wiX!Y6TdJ6e)92aZHY06uw#K@)D7@yi{z;?>Hf{{;;sT4R)y4%Fn`cZFeSDb
ziZLFYJ=eS$a;9yREt2g)Hm;l11JQG=?kS&ogqPIOYZN8_3x6?jlh4k$DTJM~n1(1G
zj4L$`=5PI^y48=BBA>Kt%I`2T4W=<pm9d7IT2e@pZ$a-4Q_&E_vE>bQ@9<>611myi
zY0SBdPd)82W$kmE>dcC0ZvpZMc^o~_)*RmFD`ClGSJuLm5$YhebF^GMQ9Na7G=bS%
zC+|D&VN>Wk$3KRO#pIC6tf9hm<)J2pOt44s{G8rOh7nTP9V3N{NOQw4UeEo*<(%d*
zM$c6=>n#WE<OP*TMC^kmsulq}dKlA!-A<)SkaLo-5QRM`s3;ZXoO_QZbpj;5uU(Uv
zDO+PS7&u%{Hin@w+{A6-4xG80msQ#c2!M8MPBmF3vOR}Cm1r<ttdfUCmB$mvGO3(C
zo}^MAljZ%qbBFCh)a;95KpG02^VC8?JY1791GoBjpA2UEq)jktMjimj``7;GrsxGC
zc9=Ki`5p~SmIa6rM&ORc&AfKg$9Hbu3d}NtzmpZ#<=<Xvv*%}!%7zL83nMSs=Cd<p
zpJ7iR#A`$C+fV^TP+E*RmgIPBLa$*>`XeGW$vDGF@+eAr4#qSC2r7Y}M8=?hh%Y2y
zvQUfb;eX&Sb0BmY1I_Y0xemEIowzBD|KfYsw{naNL|)7)AfmY-C_4GUtUc3l@D9Qq
z{!_O+D_5`%v~k${gyj!h<TNF6Flop4Z6r-6n`O}?m<ug?e_QOBhvpbyaWqu{f=!`V
zY;PjyX(4+k)PAG?)4)oIgVv%aGaY>cNH|8vr$3sMdC8^>2@#wiU*S*>?;yi}FRxjX
zXDLsSn#!o%%fpeE;sh{%XdQq%KMP!9Nw>XOv0uP7?~j1*cZ$=^Xj>%UsXz3wtKBUl
zN@wITK30%KBO+Hrt=#N1E2Z3a7w#^#pOsl0s@FyDTXMDLZ>{$eV(&i&m8?R$S6+IM
zxxm4|lk3_GiH^8;R<(l!G9U}aZ!spxUyh`Hq10HThyC@AA-_`#%Vc{peC|`W9+!%&
zXr5a>i04FV<lAMNeBQ8~N*Ce&+=jw6;!v+N`?oyyn!4VhV`xQA2oSTc8xr+tIuo^6
za`K%RqVzzKgBvS(?`+`q)=j@=_(M@&C5r~NAA|Lt?r|MR;)EAaFb((9ouDHaw`SCe
z?UA%UQa<f&2#JoifZ$>m2a0<}d8}y2l?+mwyWQ|isHUISG4`j^u1qnJ%yWoOo-BxC
zlV9-@07zy0{a<J*mj8{WVrORgAAX9Nh>L@X?SF3nz5bu$|7%73f9fCpFXV^+(Ns&H
z80<3AMXXA~Zg9eGZc=UtjGR+Q!fuhl#Z=pWh@n6NCBmAVNvr*xSEEj|y=M4cKUX|=
zhyM=x0XkoOU*aO<w9=$U5Ntu!f(Mbb1#EQ{{j*0{h#36SFo&imC&Q+PiXjCB`u+OX
zWFYi8+@SC%KJ*|QZS}2&Kq>5E-0K$o2Z1UG6o8PxfM6m)LPR_~fW5oANPd}uJ0n1j
zs`-drf#7ZZsqr9OM@<uhI^Ws^@$irlEq)%M^_xvVfkZ`uIe+aT$2fzwiCR_|{X^LX
zkgvYt=T~rG=0Y}w`k?augQ?KnM2NISJlwmwy4tF9w%hk@5fZJlff%3$kniBG5L_R@
zHi3OP;O3fKA%54fIoMGJW|0qmC~II_#oaYK1Aq~M&<tTi+X3$GeOG<9{$WQzI_NI`
zsjyftrS%_q5ZV5{rl3HMpuY&OX|J~eNNa>MD=jVdW|&s#h+ro`bzuViKy;!mw!s{M
zsQU1YpEl<3E@6N-Ew8`A`77**5PmAQ@CwQEz#0J7H+8R@7F#6DV4hYUu#ZmB++72L
zk}9Caq#z!iAOk43tKH{u;jO+FM5EXJcc;c#1AB(_UA;g)FjKSNPG}xPOadRQp=Ho2
zlHas((BN;x6Cg1lFn14+5D@{OIvSuWo6E6Bp03<D*f&J-JC3h8LH|71EpYwcIWSHj
zLqy-Vu<KLXYiJ0fE#5w0-{0!D)?j!rkSn4VD1Z1hV35Hd$`4SumS1IGjUL<~T>Xj9
zK0YG<*Vo-$y%WTzhM>WG{@?W9dOZRYay}1HlUM-Hj}|2j>?_2Z!~G*j+Z&iWaNjN&
zs^=Tjn;Vvm)tH{iS6VI16afO#d*enJ(|4)iY%f&b$Erge@DIK;>QCQaM*nBR#H@Rm
zyAA#?px?ampRmba!a+ar_uhmrKUk$Ff1zK6#-D=U-?Hi)Vuie3I0A{w0C!Fx__smq
zzMs`4q_5L?OaE&kf|=h+RY45^ULf2Q@W(3_@l7O>U0n`4n8D>6rPj~Tv!5O|8VIP>
z;5Kn@9y-wKpyiMFtNq`WS;Aa|S4gR!0uX@qOs_I^pY^_LuY9z;0MdVJHKx}=-*rcj
zl0bWq-ff~hb+3eE5O?;2d*P6neZpSAKwfRQFSjuuz`PIoR0AIzACN>-uzT}AS|}cK
zn{Aq#{Ez_N*q3lW`?=k>JlLN&p}T(ItBd=((O)7E#FCfzu+BGJB#^s4g3M3cJxt)c
zzF#e0s{Z4B)xd8Wa=;zK(=PNo;XtPV-_|dLzkn}L+$Lb1fb2C)`9sSWD)uh$_mBZ@
z?;gDK8~V$tsfEB6@jnPG{@38<H{sWPqEFwhj_&0OZA*=ItDL>-_@Eo^ALi;v-4N$u
z=)M*^eXnNgEP?u4oNlA4StxyH&Un|&t$SBpNuMV*pSag$>!Ql**>I_G&WY78?&Oe}
z5$^OFOHnZ&X$TK9VMlp=S=&=F$G4r%H@r}85j(M;qRE8IGt$WF3}wgNh0?>%oDM$#
zi>tYyQIoDL)Jgd#hDHvsLR)Mc_-B^FDlwY?v|lu4T+V$@meIkJT`rx)QNA*BEC9a_
z7n1scMRVWFh3mXiqhR!ON337MKu9jeQ;L&w<chL0r$UPTTK#X+9LwQ0nXX7y(CE8q
zz$McwyPg=^ktSSKLF0&%eG7V9?k5h?jL^DpQ?9o1ih%XGSgokiwl%hJa2YTkq_SP>
z#w2}9Hg?k{Pj^8k>if$+0b_vKRU7Ko+KrpeX~)`cbJ!*C!*prXPiU>xHE<}o(zcQc
zg06roxcK5(h{rtK1;^QOIB_z&a2?FO(G1t@>Hx>t3i5$(L0E&jd_ZMLR~N;=S<YCH
zN-vn>H=-W6$_k-*ON|U~MyMxpX4rNZwB^pGw}+Fxg2X<QQr$$f@${sExnD*r+H4`I
z8f;A-6wN7@MmgDWD#K%es{CK{bgnBC+5>gPi>>sUi#T@JXNL~7KYKqW3gx@mL<;%B
zt^d`rvNh=tE>PlP%#%Bswsw2I+~ib&k0ZAA9}E6dPqzN8y!8~?_h8V<13ztQxBB;I
zd?+6+uU6L8aGmk$eB5(^i(?2C>q1hy8on8Q*WDqta*U-OZQ9bg@*C+tgwKkfRBuW0
zo?j#L>DWS;QSaKU8di;LYicyAHl3N{R1upN@XCROx5Jn6Z)v%6%<N7_@FM$%y-AFN
zv1Q~8vYM4<|JD*L9D1mvKnOMCyks?SF)oc<!>H)?st7mjsoVJCq|B0;!n1F((<jKh
z0WOX|$UTBQ%q;~t#Dq3O`a%?M=v8@qU9Pf6b3wi7J_beg#<W>_sPq8&mTYPxIsrTF
z_+R0ZpJ9yc4ph&~`$=_jDd?q-cm{!NJwmx@rqp;z7Qvv}sW3?;)3ltQV}lynkQk;F
zp#fT?Mdvpe1geFscVb9eEj$179d-h{1aTaFwsg}8^kh;>ry{nuYM`a<(|XNcu4Mx<
zRS_-}r_wwmu|R!zcH!`T^HcN9nWN+fGLr`fIQeyT?_mo_22as|WcOO(Z7Q+fi%+Ds
z*tfftJ6ybMrpms!15H+e9cI6&Ca?Pw!kJGO_%xOb>F-!b>Gu7}6hI);<>hYgZ2U!O
zT_QJz)Hd;OJ>HZkhtIYr?Nv413}SXMf2u&VHWleK@?HdV2^$+kTe%02pMTBb^Ktwf
z@p-Y!MEJx;{k=a6p<`xwr+^j?MWe~;w|=919{YRq8{fGM3+=2b|8ei`uY!;8LZb!l
zW3pSu$96*K#XqrFCc9z%uF||cW{-*Fbx}&7VY<;~Kn{Tju|sTT;=k8vjTkS%m@hMc
z77t~{f;t=El8Ose^WW9%8tW4OW7XTdC2c~E_usBV2)ha})h2B0f>l5bHBNitxSGu7
zm>Wf(@HZ1?w>;CIC5Y1hEF5@B))hW_*_8-N#SWf*plsA{9b-Oeyon2rdNNQAr5ufz
z|4q{Ox^X)(-<h>LXqQ#<Q-v%i#rPQ=f3~msh#?inX;oWZEQ>FXa6$`O51+X_8bg>?
z<rZaXn2PM(<u0u)MLNdfXR}<j)>7oQZ?u+i$h|4;@y+{CHz{d|b(|cV#YeN2Y_lO>
zyN$Ox@%V1xO)o0Gm{;Tvo_)@F^0>OXoUj#4JZSzMa0=|z=TKgLNh1fzG!cZyaV1%X
zX|9}-x)W@$w}R!X4bi|;s)s?J5qL1AIb28g*3PIsbW@?fOVZDeCN|A*yKfdc9+r8i
zavZnTh(h)G!HTH9Hsv@|2!<_TwlclWTV4G)-z3^HW=B_%F;qG0))^e+_(aY4nxynR
zmeDYHaAl1CG_k;bWJqcSFvus52!zV2tL2y}cV$1N0V%d?x5)HAcwnehP1xnbho$T<
zMk8~}Ad^NAut1#CDYez@He$OkB)!n{F|W8t4}zlMW^@v<`L#jBlFu)_;SN$xx~W9u
z-FvUFwcLUS?H;dac=NZ487~@2SpmB}xHrV3AzJ*mnCfkgwXk9h&B|ILp3)&}iz%y^
zLvsUoj~$|hFwjR9nq=6WcPo=p3z#>BGd_ejDF<nRTSJ+{aQ-|YM$7j2B$g)soQ0Ca
zm5k!&pc#2CO!ZrtIoSEBie<;R@(ECrv9;i~Hb#05u6lhFp=`yGm;d}(VxfIS!Tm~M
z89%nYu}Ig7H+oPtYJfa95OS|Vhg`b-bHb%@>P_9Rw2>==^rYgUxw+vql_umDKJ?%-
zgX9o5&O-$6_3mkhXW-HWAzm9JZC{+eYxqv2Ch-p;+<CK4z?!1ggvt_+Fs@E2N{hw|
zd|;h~_RBVs!n&1(d4O1aK*mrsQ`g~GO7>oeVfN(1#;{UwjE0J4si1<-YUV`CWQ#=S
z(DmK;;>dsfbyM2P7oTYTvtOW(CqY#`sBk2)pF|+BbNI2_+pSX_o6&oB$aP9Q1zann
zP+8?-y$9{2HWx3sb6*7WI4#V>a-wy_oC-REv7$`)COyf-6pyNRFh7ugj6tRT0KNGl
zZfLoV-L)wUrdfA<EZ*-qS_a|o2gJ15PDL-{PI`?|2EG_`Htsf*k?rIuydvn9VtnzV
zKpj!zE^nQ6L!RrH0qm>m`veuH?1_JTy-!=5j8k`(8YZ#;cHG1Rtu|%`vaRwD$Tl`V
zZY>&u7Eoc;T^hyPpLWueFa2v0ISwBjeA&GZ9wr20%tj#wI~buuRdVYnQQ(k@f{<%t
z6rl>d$&eD2e;<#xkbpo<3uF5NcoCyse>K<vf~`>kyxtYl^V(SQVFb<~@6XGT^t%1&
z<;nJ)GM_HHZQ(qbXGJ6&5cF<wUs>y3O$g2{F3OQ*FWk56m03RQ$wCgZJ9sib<%0u{
zrPwr4{ET#0_C=kEYYLi&#00Pvb6WL3kX;u-$(xca*=aDOG5o9EOvptq1YB;~Rr%Ma
zBFwBJ_mSwJW)y0VoE=!2*2i86+d877&jgN2$A?)aFRAADxd&aVR@)-`BFAMnHp3(b
z&r<j9jxcJ-5P%1Kb^cd3ri3#xL>i3#W`J9DEFUziA+7wI3;+V8Iom2qCh974s6DDn
zz_k8vWpV$eOBPEP+Oo`_jcX@Dz_qMJe#%H4;^1^sKD&0b{>kAh32o5Ffi<(-9rg|C
zvL~I`)0Jh!+7_nuoE7>hk`F%fm9aV*tKQL{mWMxa%jNx%7xGMt9@~QDW4b*+^Q3&y
zn=Nx-+3T7Z<{BaBVi(NkK678i6li)e<M*^Sma9wEREH(AZ%tX|vuo&hb4o12AzElo
z{6=3wpBG?W8pxV@#D1@D^e_6_bX7Kq0$i#8T69s?MZ`VR1x#GQ`&H_E^D@^>|4vXW
z4R3iSU2?)r4pbMs8B8%V*=^q|);`7$kI|VMa>YQJLVLwjROE_B290M@$LMsX@Gxe^
z>S)$VUYC|=^z!=Y-}4TdmV>^q#5>PQb$9gPIqh^P01(^DH*YB-=bT*hd_49SP5J#E
zpzr96kP@;Qr@k3pSBLQVM#UMEOjKT)Od}c|Yi~*+@8Aq^EqX=r{@}@Qb8&580hqO9
zYzK*w;yOQBv$~8Kv(a(Wpdj6@TV12IjZDp!2T`%7E)MAR=Adw)^F`WzWCO5W1~%2V
zzIDn8xvs*ze0-$%2F-i5Rwh2we0&m6Jq9R`x#j~S0v;aQj-5Y5OOYE~fO8m!jSE;@
zFTP+MA!{`|#Mja>g>WM^Mi%%xP6EP^xu`5s(iGaRmf|9fX*ZT=<+*nzuS3m;lNGjx
zk1O>|U?x{e6boK3H+H`ICy?98egK5nJ4+GMH1jgZ0CtK($njjLDS};VU2k>vB0@j2
zNC#jUe7Vg=^tmpiKW+&}s7f}e-F!7}r-I+;-}3>_#Zy@#VW@<}7Jr+md6TP-cDAud
zq&}k+hqU9<xh^KwOC#Zx<dLt}e+SDCr|MY6x-T=Tz@J$(t<KW$ZxK`57e_P;n?Z2M
zm(3@$dQT!=iwKdsEo;ciO5v!xPaLSDHO_OVi(oYxq~4oJXRRgBx72(1{~0X}g%@;b
zH;6KZEBvXl?uWU^I?81$;ZDvCNa%8s_ObS8fYpLJ?;EmO?$8t-GcR;1KkQdB+L0*G
zqEGm%AvVQ>M7a!Un)EUhHLN+`MYv6)n&eaOOayiFvDPD?m9rL3m(c+KU3i#?_5P&e
zJIAS|bu?Nv`B6xk%ULhpfkxYgAzNB_3B|gMLwTMzv~$ZJ{HN}L+6Nu?M})cdF!G_I
zisQJQKzdQ|#i-u;9!y<oZK^{Y3Z#M_QilBk1PYGNC+IqpbNl$Cno7iy&vXfss&sN*
zOf?_=3a%LVwdm{te&2tmyu@<K2i_0fgb%OQY^eAh>}GX+&%t#bO#xMiy$f9Yv(aT{
zmR3`orc|vaoi#@N(QD`&();{o<AV3kFwqKR^gMhqPVdflgtH_^Kvf`@L6;(Tu5iIf
zlZ{rf!giaqQw}jP4^Nq)9ClR{FfSw{UYwujK{&v?0=4``!*;3}8K>;h_0)(CXl{;O
zjXDbty$D$_s)$*f2A7mYj3g!^9s^45uQq{^Poz&m2317TGWN<GL+#ZQ$?tt8PXw-H
z*{cxs3x!+Q(bk!Yi_6Z~BZsQf96i&?k|o4_b^Ft+_#PPhXT6K)E?i~k4n+AE8@d*m
zg^tOI%DvZ9YO$WVzC2bUqoPErAyvwHrl2fW+2>k^Qs~TRsfp~9efEFK^VFKYPar<F
zB~y_@+lym2f={41N6V2J0f$?hZY9l&qzO6<FZkDbanvTU(2rJN4zp+Ym=T?42Mo@B
z)I}LN;cn-N>QAP>6R+vt!nNjf41kevo8B^fei$ijRhy@&!CH7GJ^en@Y#$Xl##xe6
zDs+-~3W6GGU+L{Gj=P?31@87lURav#jzaqMC{9_5F?Q|;NHp7U?b3cvw7C7yC4~V-
zJ2pNTmKUeYR*-jLMc(qtD*pO10Vlg4HOn*h3)=^|(Po~Wn~#4-MpP=j+i6&?^*b$n
z&>0R}^PaJmDa&^>M$^X!Fs({!gJuEq*HkX|Z_SV`KK|8Klwks$@<iL$h402I=*Oa)
z^gaA$<c>5XX*Q@N72e6473lM9vwuyXS76=+tIg*;Pq`xnsWtcOyBCK4k>Nal*G+QJ
z@{kCx4@Ljc^_z7nfscT&`_Gdb=xrjvJ`?swL)otRo>U^RYr@=2v|>0zgY7&p*$NNM
zR}Nk?OD$Db0;--6y3=RpE!vCn@M$H9ylpijE8axHty;Lp!ye<pNm!abStc{i%A1?~
zwV+}1UA%jE)RG(#So&YIIOcO>-+>hfnEVyk%(m3l-UaXr%<m+QJ~{2g!(UzIKJXx;
z#hxv}p@f8wiYDB3ZBM<!1<R2Gt)#?_jPyR9XrbDj9MdKjbdFV@45j#PnX`s_)NpZ8
z`_Ti>)A`f-?`^x_5!*OFg~`0!h>_9h^?6=TwcrjYiBv(}{_TBg`8_gIt#4x;-L>Cw
z2N#hc=fa^w0%f--Z4tT%k}t`$+`k#k$cG(i!hR3p)r^a(Xt!8Wt2=cdgskW*|0R~C
zqaYXL$J910&%i&itTxdRJr9*4AEfxuy3->LczJd-G9qEGtsP=8oh$)*u@?n`jeG19
zy2dnduolUsKj!G2O=QZjikmLyd>BrF$-V9kJ|i{NRfrEzQJ;5(vX%dlB3eFH_$p)c
zXLp_(u{~C@!Vy`gozPrLtq7E)aP}V`zLk0(9Ymg_`8;D;`cAudq#s31@N2NI>UJIm
zS8u*=%L;#zn0F7;SM{Kc$Z%VOU(I#emPXS*VQLg#m#fHt)m;9`;H#Ct_hR%ub*4%C
zJm8t%)H8N^jOU6lyHq|tdzF{O_<)UiOJ6Fvvu6{~*tUoeRt~#et?ic%#y@{lg2izU
zZ`HHj%aQyy9vdEch*V6bdBI5L4{P-D{X@$C7z_5Jbo^BC6@6ah9QC58?!wr`{#gC9
z<3M!dp{TS_7Jo#qCBN^G$xKU(d#r5nFB<6^5%4}{cEIZ$t`&2$lO+8bbFg=LN<uA0
zDft3!l3`R<`AXWliYL-I{NJyVm&bj!`T3Sw;a(S;spqVs57=d~l$4JYZ;0)>WsVk^
z{;<<DR!2_W+p=6pE3;7S;}r~9RG98S6y=*$-zf7PQZitlj8`u0ZapHVeQhL82;dlS
zZSviA!6}h7vM4C!9UQHZxKNfWbkJzm?^)0~RbPr+0)$<ouPMsfx>yU!@8HuIge49G
zXZnNBvAH&JSQZlHtPetC)h==96cb+;-@ji2cKd0(ARa1`e(d;tm`R|pXcC;FZb&cZ
zFO$wUN9rkiMLo3QlMC^QLJ4F6JB6<7GmP0~X`BAF?!G21&rdH~h9uZ2AAhK0HE`6j
zerL^qJQ`~hcunG<`@M2*Hs@PuHM#ToahdEwN7S40js9*6&O1_OQF=Yf#6T?%iSs7v
zJvQC=nJ9H)(}p-JxM%g+4VWhwYi-Y3)N&w(n(7q7^l1@h*&fGGNAjI48Pl5(2PTVV
zEt_97RpR}Ths%jxXpuaRxRIQtx=FIJ*g51~V6)Uasz$EWw{q2H${v7vpR(I6rrO=Z
z0?hI~3L=I~3t{?1M^S*BpxaMIpqdn}mkBUahj}ac#2Q7TzQW;a-)=HFxB^+en}&Wn
zK-9arnSJT>QRm7q%`p#l^3lVo0yY7m!&K!xucB><tN}kuMuO;gEE`RC)gdh7xbwcC
zlUB*T<Ku?NuIM{J3-_KZp1s&8Q$F!O*><xAy`vq5C43iq)&@i6FOKKx>><Ha27z<9
z--S~&3@sI)kcDT!D67!(=sc96g(Vl@iBzNFHQk(|8FV-HyfqZ(WY%md<OOeRA5wlf
z^O#gw$LhF1Nu+vizu|NKf%MPg+&YD4B=h>850a;;i3r$pzF$iL^ay%sV4?@f=a8MD
z8ArIwGL0dK#jGhR;>9hUPDlM5X<$ECpy6tidm8egQ>mLetdgpm8q4^@`4wi6r8w09
z<`qSqp)sqVf5E{rSFIe_-1{yiu+j$~*J6b#<^@iVN-cP1GBj)&UYXOd4ILmP377`g
zgz<+Hjpsb29n1hu>&tI*P3_jYKP*=xL+NJKCT3tmo(3w~){O}wqZGBG8c0?P?W(Rg
zpBNFCqRdI{-N;Gx$_YhH2N5vPeJ73}M6e&3@B(pA5Ozf@%WruNh#_STiZTp180WU;
zqj0L@$u#xBE(xx5+6g{b-|S^|vm7iHRdh>oYO;-HUAAkps+DW(7MC&45kcMdEXYQy
zU4ImVV8%_$+;^>MtJG~7t2y<92DBzNsoKDCgy;^@Ql~M(>VGenXWH!E3*C;s)gK1$
z!e66YS^?eoxpF(UB8@)p4gG%X=y8x1tkY2+A4);AW$*U%{`=52Cy}lNEbDj<8Q@Qw
z(ecrZ#zRQDYKCwIA_5o7Qs<;CT(d_EhN4<)!uy#-VdOCoBm)6rNunL&2f~DtDU9n@
z)rGJMN;=&oTb&iL8-mEh`_20{%-)hl4PX%G)X`qo6>jKje8XpatkgjIrl@`ci(lAR
z<(=ud*f42GOD8ISIA^liOr)$Y{R5=a5I2-MRvyjTSW$UY5VUp+0O+3V*W*?TjhdRi
z=H6<XJ!`5$pV7W4WzYLc2LGN0&nP3}v}4x*spV&*YWJ;PYmsAm3^kfrCvzH)mlYt>
zGjCXbZzgk`<|m+>TMeBr)>z3yQFCTNXavXmXGv!Zxv&;>fh#~Dh0@~fdCr&Y|CLss
zu2$*t0^G*)n_%ONTP}UjA|-le8N7P12}<l(y}-Tm+96VT_Y(qJGV?P4{Cf_vHu-z4
zF0v_R&d+E2Z1`Q|&V|5HoZ=y4!zV2Md+X8LF9jF#0A}$!C#`V8<w86}K3QU9ziPW+
zU#beIKwr|LZe&IXzuX~W3J<!|pBY~T3^UCsH6o&_Mohc&n`_5}F@Xpb(bux*D-^AK
zYu&_H%V;=km1po_yi)Q0yTh#5t9L>m7U+n;^#z}~h0QkyOhi0O1Y)>V+ZjEZj>5Ix
ztD6mIj&z*ND+VpKOKtkJ9xeZZH!GDa<gL6@+=s?{$=XSeN|J~A%S+zExc1h1S7$LD
z>7qqu-DD>^OCmM+a1SOhely&$uqH-}sIzDIv>-sMDQdDJp#zV_`l;DC{syz+`X^Fu
z0W7D8kGu8NgbT-zExtc2_xD>#`cq%)R0DaqJHvE$o8Y)gBhO%W9QV&l1hTiG!hMK)
z(SM~ytiry>i!JXyldq?n;_JSdgwNUYsf`9H%1ZmTM=cH@(p>I&L+d9T=j0nkla&k|
zx5+EsaN0zG4_q>?7tzH^xbuz2CJN>$Znt$%p|Ed4k9>$mn@9Zk)U8IJ2j_HDodC#1
zfsjxWxExn5jdrUF&6(jDVbPAblF9BUv8EdE;$yvdxXAKz7ZrCB4EnhFijeyfDsTM9
z=)Ws8&+W21q20$wBE&Z(&qnv!;XJce^V$_|;*{N@`_mS^gu#NyHw9D+hD1z{w}RF`
zn_fH4lk76s?L}lupi0&dN47AhNiy%7b~k}S>IIlS5hdQVi_5G1&1T;DKdC@|jR=v_
zJD+&#iRy#_J)}Y4-H`zKIU>M9iZsbOH2S4>v#PVv>l+8L2m9p{p+9)EcAAE@?M;&u
zVS-UyyX6ecTo>ZR$WUH1<mAguZHk=aoR=$;fPfxTXL<hX{Xf~){OZ>8@}lM<=sRXP
z&#h(wXi!&j!iO<WPekXH<e4fd3$2@<HaB=!xBr3~<9Q=5{9r8(tj#-(Jt6F&@&MKm
zKLHLqyt}oZBYFhUjpfU2G*6A%&rt*Ofp%J58~mVrH3}nE9%tD*HWf$YZgeK|Hq8ue
z?TWOXC)Cn7TYs4f=*v4sa}$cEj_1=xzACW6HjgmG4@s?qm3P=C5n+u$Xir#(+7J%(
zo_pE|=$z)i54$awzuvp{=Az@|$%B(jdT0t)nYTwnJDKDw#RMQ8C8i}y<;bSGY1(Xt
zpGu-Dk1dRn9N%^PEZkmr8Fk)y&20-xXSpz=aZ9S9#%J$+!=}LyJj2K+rw+}m5%1BF
zn_>Rf{khJ=ovbW@-wk+6R)cnT1MC19qrL57jMdH8x$c=-^n>SN$%w!N!ps_L6#B?^
zh9oQyLYqFv4SY8Vhae~ffL4<U@+_++8gZ#?ET$@pB)XSOnZtf!eOXEx%@`*^CNvw~
zxZ$-!oU^^?%Dwgi0;J(s#!5;4U5MT%%w+q(e7nET6YQibStUTg%=BtH6GF!U^oa#h
zub{;+t2VmZm8>w_=RR7JszRdi@Np)eSJ9^BgAeY^E<DK%6X>mN;pWIlp+F6TGO~ul
zW&5IcM}`3^<d}N4O|9?!+8LymIEkTIGV|e~>LAca?TPBzTNYgnA>;rga9IMrd&da}
z789=oq%>TlNI36S{gqzutjviH5k1)*y2jF0v>27ocwgNiayA))3HrM_Q)RrIujV}h
zR2R}H3=28Gc2@G<vE6&Dt$Dm3u@)dn*r3iWIzNZ~kUIl+^0WTM=CQYMf6lAxhnO)e
zcppKk;ei=lwX_J_PU;e9bp|UaG~Cl$5m##rOKR&K_oDj%)I9Ji7T*zhI-|XG^v6e#
zR%gQZu_kGyaLngFdtn@1;zE#I&Nn?$Qa2b*Si4llB{8b-mK$Alu>Pl{*(QUbjoo^b
zdApMSpv@p4cWr9@we^+bYA!AT7Nv)h-r-4Krm$og$wJRL7Lb%qT(Amb?6@#h?QJ^V
zc`5VmtYGL#qA1+z@T(!ow5PHkHgHdEUxoeZD>6t&6CW;-A9M#QhfTaEeLeZ-3-zUz
zjF(lb-?l@EApt^koVa?TE0oevXqVW%Je|`g<ktS((MKv9koD0J9+v%q6noY!tVhy3
zY;SzQhS?gaY#lz5ENOV{?!XfFTxNA7QN91rVnDZtz0k02K5bDt@`H=J30n+`fE2Uk
zMwPxjXWwXR2|@lWeGp+^3{7(8QL{2nac`bDn~!{uIe^8n63iK&2k1mHL=oFI{JcAQ
zln|$4N`yy8kttMRMK$4D_UQ;`&|0<`U^IWN2h);rX=p?%?Uod>^ht@_!R&pVc8{Jl
zI4*-jP_-YqsT{?(EAaRXyDj+Y){3t3X_V4-6f^;u-_@pEaZkXFBv&4i0qfXPY(s5%
z5il`B^WR5g693W5UqKXhA)}<u_WsjF&l33*bP3?od6I_{qAp{Z>`}ci2*T>!6Qv>V
zHf1)5>Tvu{-$XnhI+SfOhiYlIw#T~yygdtnPmh<XVxfjrE^DLIT2}M9sYsm?z2x@!
zktvoP#@xHzClw^devpAV=IVU4hrq?M@+yUPS2^}X3iZwW7Gvb3xsv{>IE)gVC&n3A
z?&+hxV8QNq7$Z-p1}RI@cbq<bmt|SLJQJl6#1yuOUDD|$J&g*QX_Y4&zH4FzzRZFf
z5+bRxTSqq2*%+$Td!50m5ges#ONXG=&c)i04m{};ttaA;)l`HGcVPJs6s{iFZ5w#$
zh!zD_H-}UQ5)(^V<^aDEu!gFMJWpq|&S-;r4)EZ`k4_X8<WFK|$+y+D$a`XW2_&+%
z5=vyw56NAQ&v4%LO2BnRIa9;$#mp+vAVzeWPH3w9yg0CrT_J~@xCcO>|5L|fp?Smq
z%jD$=$EZZ~t!n;cC4%;S9EIbN)=$aEA{|g3d5+R0aG^A7AlqJX|HhPcsUoS@FAiH)
z2)E{4dyi`CAYwQDR(p|So!+&uxXm?e{TIL6Fm-&<P}P53mo(J&yxpI;t+M{?a3~>}
zjXaCrd%32=&Vx$0Lfc@TDFAlHlQ;L?-A(dp7C)m-GDy~N^LX~pyX(V@^?08OK58Y?
z#x=f3f?GoACbBa)uU47eGEeVYy(W7N_)dR!=uxfWS_EPh1^toC?q+@ioHxqHd-*n*
z2nf!!!5P>?CVhf}e8;b6XC2GBuIY^C9=~;0SlB$%fHDUQx6111N8up&N(aO?T9e2>
zp<;7iwWFrhaH9*nNw$Jrw02yYAzsbtG`=beO;@#p?*pZFI^!32ecr%VzPKOY$UN)!
zESJ|-uy*tPYkBO(!!C}fkx)xC=_Tpd$IAqiCqQCT+qFW=#Z~yO!rp6q1U+Xvrl#?u
z%Drw+fK7fa$r`fwa?PN|F!0HY$2K?;z^p$G1T;MSQgT4a%}-oOE{K5vnXnQLJs?0;
zM?Kc$Hfi+VX0GSXmM>kAkpd*HbO8(NzU-(k%cR{DABA1Geo02S%-Yi~Cvof~BFGt&
zHN`>|z1tM8ffTnidA&BHX;{Xply;Q96UUa{4~`Kg9+SjUTl1(gP)qoXTT4*Sj+KN~
zaW@>*l^(>9*5?fO<elJa+qCCE6wiUInDNdt35VA<+XOmZb!k1%8gIZV8sGk8IVM~4
zgC@!{pXYQ+d71DvZ-urWGntDR$+yXkCBhp)cL!QX(ygH*Z1KIMIuT9aJMw8HE!`wh
znmy;XT!^LIPW{5TnEXA&spYAKWRk<V(x0%UBoUN`Xq_U7Wld6*u!F4fPdvi?>@=A|
z;2B3#>vr4aG4_+8DJq<24<8KIrHaH*){!%+i!wg%NEiRp#)-ctC_E#D+4bVpheOxM
zMj@j5yJGTQ$G11FQ!cGp+w<0QV8#Rmk6Es-laud+2|B~m&Um%wA?&q>RfH*aMV3LU
z)nrhH?iQ_S$<8U+qIHQD9;OzW`%dec+wG>)tbA*6YO+4P*=|VzJkrk$2L;TMQM;rw
zJc&FnW}%88y74r<y6XDto5eNW)M)>nB{CI6^J!IB%8Zmi*;p)?+l;A!0w12S(C`bb
zQuo*KWpOZ@{1E(Rn!YA!nr5oNur44Uj3mm8VX0!P2(~;u2osg4loh$g1j14vvrJ8h
z6W5b$>p^z3#54mWkg6#*a7GIWOCHh6-BzDs1iYkFCTV+ER6+utPMW|bm(>g3i^%k>
z?IUBGU5fx@QQE}tp<<}H*<+GG*I%WnieP+P=)NujCcZMWFc<3;BImvmDK(M{v&rc*
znlB6Im-{YSjT`luo=b5`igH+1-hO@6nz??Ho4kC)gH<~MIo}gqYRki<Fl2iYrI?d<
z-Nw`^U$7>Q`?aBtbdjyAVJW493FS_@kHfj6zMz-xKVskD_d&q9z<?CI6g>#|{wh-Q
zUX1z)6}e(Rp(v3-)FMd-j{_d9zsZ;6^B2o1eoy?`fnE+%o97Ki-L4wNt0MD@@i!Lb
zVe}?Pp(4;uk!Stp?DtIPJ|N+icvRWQK7q)S#JZ;T%F?L10!@;|_wV<x9c&KX_3Bez
z3+(Tf#wD|s3(D6YR~3a@u<GXLPTX3g0urR^M+F^^hVb?4MOe=mBkH14fvgB+fynUn
z7HDlWF;sYgoxqw!i8OgtBsL@SZdLMM5rU%4D#$vDim@z?<Cy}D!<AG7klUEedyA>V
zz`x~<AK)d{57vW8qh*hDMxv^93qEO1Cnw>;Uu=hYuEmWigC@1>JI+jdthCj~&g}{F
zf2M1RM3o1S2vf_#+M*RY7%r`-y_+Dd-9&j1^PmrAPAPgiT%N>A@Z6&J_6$t^b;^6w
zP=lkF9t|d3qJxTNcCTI_d%Jm3cbw?w2MTg|EvxAZOFF6{&cmy-V~vd)2q~;9JedAk
zE`rqcfh|J)#Xj<a?GB(OCr8~50xEK$b(eDV?!a5~oeR1d^AU+Oo{py{cIed^YAaWj
zB#@@k7Rg%0ByYLgEJ>iWz{;19HBoN3u0C*X_l-G_In6%En%{SZ+HhM)V|g5srJ#Gh
zLTSG_sV1dREi?Ir<YWTd{a>(Dw*L)FW#jn&G-S*~oGcvd|HGpG|FKjqR%Vv}Yblv4
zqzdZaMY=#GA<TIC0!d0}RRzz49WO*AA^{YrZcOm=Kne77e35wSZkTY^KMP(d)Uw4t
zWv`vDy}Mpk-K`C3j;6O6Tr8(oTr^eDI2*i5D^$|yyfEhzp!?uaASNo!!2$Y=00yin
zNPUy#fKKj#pL`%SKCBJ!SfR)t23-jFppG5=28C5z4plUivZE8I;5$%ap+Y1BMMkVK
zNQkJfQREN|@Cl*q{S^UK-Fy@nXeZ#hiTv!YIeRUh)hwN#Mz97g2Owe+l7X$>!2l&F
zEF4>KQ4rOF8dt`l=~y)d@qG3P7zC&14|SNAQP-Lpxk94&urNplwEcl7q0LAqdr-c;
z8dm|jbXd4+ghq%TOUyC|(LP^}5m<fji?x28pS}$$Ty;b$9EdJpynQvs;z?P5+y$Ia
zh@C6=B{db`3ZDLWziibXPJ7T_ZXBXu%m=rQ-a<d}KvF+Q;p!>{C`zgY*w`lUjXrz>
z0T&N+U3GbNTOu@z?@8o~v%rpsunyt<ScOU4u-y>hfCr^afiqb@KbKR(kWilXj=PRN
zd_NbckCc;>_~ZB$W-2g{P-hq5uL^mL*ig;w><-Mc*Jf@aMFNRDfE6z1!p4a@?24@%
zkq+6`I(!M$XWA${*c)VHNIM94zkXdE6c&^d!~k#gl|XOer8`%@Z&JM8o$R+SzZ?~1
z{<JJCBcKib>I8D{5X2=6SZCgr5s>#Q`?U!oKqO$*mO?X!VhSq`@-yxWI%xf~#=#L6
z>ID+h&oK}L)ZfeRS1V)JWD+C9?fHH9^XVwqP{EyfaiQId`;(QJD&_&=<Ng*L!0R2T
zPXsnF3W?UE`S#5d1%dXXj=tw-h@QYsl|PneXeW0L0Pp;D`MmOjhk(4>ZC_I0hztQ<
z|3z+t)<Cj$c?5s|O?|(O|IO;~GkN!e`s%$*aC370nr`?l{6)lw^yl&X9^O4$o_-el
ztSHNZ7Wiqc<iDD2s~D)-w+;U_Mb|0dU?B==a{2m1+NXc8mrrko*jk>riBtL6>*$4;
zLXHi#$k*QMB|sxW1cUowxG}A$!m}lSPs8Q=DbeQ;e!QhSj#s0))LR^;qyh^roL6Az
zn?~Q&2oxY3o5uP($m>Ug1$7`=sBI_Y&vE?@Y@=Y_SNn;E3=x>$o4>mv3p1Ao@hkQg
zK!p1iQ~V3@W($;z`(XzJf&mt|_dAIO9EkLrj>L!*{I!YJMr^0s?{|f}a(;3x$UR{F
zv#tX4fEoGYMh7AgKRGS&WwQ&`>6!nt9(4=x7KY&t1?Dk0=lAA^pP7RDLD7C`r?{Ja
z=6AJwiwbuF-3THRTU15B*XZ|Kb>E~Nxe{dNB48ctq6)|e8Z|@NK5`WcPU^d-r*kLp
zAxW9rlw)qB9rdN<8n!=gd@g`F_3-I39c;f1@yK4evjo*QJlh%V-=Zs@p!ekECcqzO
z+m$xc>n4J%LTvq;tih7AO@LxyoRs6~%zs~XMV<SD$K6xW{vs!V{ag43E8k?BEUC~>
zWtDoI#Y9bEtxn*nJ=6J_J5)4EN$SP|QHz~?&<5nYIsfIsF9z$)1Q_iylXiL2{#&kD
z7E^XsAA9Rd;JZiKolqw7zFf*^M{l+?L<h+BD?uN8iuip)$-OgTaNin)i6EdIVcQ%g
z+@yPR@sYTpG%!()Xtzq+A_q2+OAoZV#N0cK&IL-;$#eKP=w!Ri_)G5?Vj@?mM3ovm
z0Gg7HZ80<yyROT0RYY34g*{Y?m5ZlHnaUIz2!+iGeH@`u`PFGH5dPVet$8Q$;!Gjt
zuBZhez@-o`lbug<3Np)`t2JI#%pIKi!T$P_Z&9bo-shE*n0Nw9Nw*`&rk)<8GY?hL
z-7zY*82!jjvC}gMdh^ES6w#MFtKMy(_L^~!fI|5WYV=`nk9U3_VUz%8T+Islr19=P
zkANRxg$#|sGcRUmrRU-wt82bXJ|us1l_pU(f>8G`tm)BQv1<}#8M_(NiC6Tl^0Lza
z|1*bh;<%w*BMAraTG3+YBhEvAO`7VyS(LQQhX~lYvE1*DSW}~H6V}cN`QBOki4?uO
zykKuf(Lq4FxoNfh_VM<zV9Oe7IbZ20#p&i&L|uD+*ULYl^~xf%gZA1st{xYrd^2Ko
zkWVZwk7=^$+5&@OUx4M?#C*8KJPlj+<?opY50shn)tq3~aF#$S!Bh197R*k0F&wlb
z17hrF@2*I4oK`%N?=KwVw!iMWQxZ;SPZWQ){n`u7Egb|>>@-_#Inq8nbH>S-?@El<
z+1uIt_lJq6y!_!kV0NvLjBcu)YSQ8LgUgYzYW*m2Kb+I*RLx5blaI5<?Qo_(WrWAV
zE{{Ib8W*nqeG$h>VyJbUvfP><Kfbm)aRs#DM;GGk1Z<tzrD#p4U3aH{4ZFVD-Bqn|
zfC4E`*9xCpT!-a94~kw#hx*#-s~eKn^Ah{>5~thEg2nJ%KsoKnfR&31l6d}q0Dh)1
z6Ey>19vsz^M!sx;lp5>fk@{ibg@NL(ZHoPJWcwdAE-+PlPbFn8-0+R_CJ#h&WUf+r
zci4w2b~`7ZnbAS6GV}E?omSvrYhDlchSAd)4%22d<(nk`d<Mt<PbB`OxrTohB_-C%
zPq#xcF{C<3F?vr|4jD43);XM$t+X|z52}t(nAUhwie~ubpVmfLW3F7u*dENR>G3#A
zZCYsrcD!SdPy6!wTpcw?*9gSWkopJ{h-FtA>YPOvAZr-_AkDzhK?owv>;YW(&iie4
z=!m}LYO;k9JXSnriT@s$aR;B~)5#NTSz3FWyCbc$ellAl1=~}rIQ?WEIsc+%2|rc^
zo`J-BpRpIa(J~mf`pb4hgkx7zWTkS!QQj|Df#=1DkHtisHWb*d8&elB;h1ssK!5@u
z=9>O{`S-l)+By}X(T?P=UG&+UTuV3`6KP4i9N*~#i>z6ID|Vg5*jjI<XCNu~AxJn@
zgMwBr=BhV7mXyQ}Dap)RhiGY?>F4rEU|W!MRr-?EE%%s9^Q}{+1@G?(x^HC7F46_B
zG&9WYT>9B~Rk@j5T>jYS5)PnE8Tbk6l=pEd?_3$yK60u9Hg0K))dwqeFR407pMc$l
zML^t$zm71C;VFOFKg<-d8&8)h-k!QgNob1b4(L@f4UBtsNd3%(SWzkt26C#1FOS-`
zgj6#VmX8|Ws+m70cbstH1i?M0sZT?d8s^2JNDX-Fc8nmgBaX%9_VjS7){Y`Zap=V~
z^*b?of)TN36akuG^>%5)y`Y;aVBhtXazur)pZvr%Q4v$vLsn=F)*Y|G(&H*=%~*L`
z#ng>?RIN+Cd6K6dK3LLT!Coi|x4dW_m8Y2?6xw?Hu^Qa9qgVq|Xm%6cObG2h$r-(s
zZdN4^vVQycD^zPRR5g2-yotKLp(DgPVd85qDOERjpl>@K1p3i~&JCqyny2PCLIE|u
z#w}gXh;sCf$m_d5fw1Z$sTix0+{n{rUIc+Oumc6YZKtm)b5xegwPWktZTmiRxF|#X
z8Y$`*Z0q4qTVN(-cjYp_g$)e^)?`WY9mP!t0!<=?@^}pE_0Gj0A_shA%Q?5!>Zckz
zh90vSLlq=j*h>om0n7ax<8AbnLJj)d16YnCketZsIqh;%NldR({0()N%@B49b6<Ai
zrltF`ht8ewD*Iu*aI+I|D2$8vKj-m{H%htBg#=F}o<fIVDt3@mp6Nv8LMYuGg?J|v
ztNJ8T1#Tn}X>-HI_KOl@7eIN01ecKuVEIP1sk8JC$YTmo{|q-Js^=?feGB<T!uTMc
zAjj{vs*xf@4JC6xq4@1_d-{?DkT5ALu-0r$BJz@?QqBKi>>Of5f!1_er)=A{ZQHhO
z+qP}nw(FE_+qTuG^U_HV?n`>O=X>;LeQPCd1odu`1!06+&^A06_Q*1VSbs~$6oC8y
z?K(|KlL<qnR}j=`tZKUVoLDir#5Lp!eCPt&^;#`*y~|Deo;6_B!G->0aHuG=DlN+h
z8C4_m$jS9!#0uq>)0%(8MGl)kU5~?Z8dhUh#9y$xVEH;LYVs1~!!osHE_xGna>W_o
z^>t{o({Z|jVv|eELX)kZF(DwMFwEhAZiwky_T;viux<}6xz);UGF4JVYuSQJ+gu98
zMjKD)_pGly8cM>AZhP9zPA&H=Z5t&r-XQJ%Qub3B^so;&mq)07Gv}FZ2*JKAK&ah*
z-EE=0&AbZEFO79N^T259#8~P!ttc5LP(N-*b?+ieVuZ}T*Tk2;x1^2!^hhBp0v<yj
z2#f?>F%a6XkD&fiq&9tC6YIV+I(Y(B{H6BjOA;6Gv7lp;M(iW^xt~YiwR;9WZP^3?
z`7?7Y=TvB(u@FF@W<IR@yMqCm3h0o=lHHMWh<(Mi=O{EjnD@*R5@;rZMcl$0DZ_x2
zeqfx}b5jl0Ck8XFJZ;No5`TbXmGVQ;khQorQj8+132WZOa(nzz$ML&AMRf`#T*wPY
zdJfJLD1|Pp?u+Wq>2-Aa0hGGq86pzG!Rew?#Rz`C>j__HXtQ+d_7|<v?=3UgO)4M%
zgyPmV2*ZlcYzN-wxQBOLVwZ;~Vj{<EaMtjN;InN9(H?PvcSSw)X+7%)TCe|A)smD;
zW-NR7=S4C5hpU5HCNwzo{W<p%p0es!o`aA;l_YSwlKr4P75C5`)-Kfc$HSC3Fj5;!
zE4w@9RAvdd$xa)3pvXUf#E1rb<wDujIB*ytUsDlLg~^}0rD)Uc2<LhMD=+2r;&G2G
zYg$}$e~uB~w6t#WH-jee^nm7W8$bJ;r|B560W39At`kMH+&h0m3p!fFOS^RyuZWR_
zNc2#fFMIC$sOY<rDdcJQc$>ll(tBk=ENf58ddDTtI?s}~^=85#{=4Rj1<+sHcPztK
zz$|>=Hz{Hjb;vJCPs_6~xq<IIf7j&3fCzWB|G}Evzz~qT!`4wr^-buOXFV!3&m?TB
z>Nd7aOTqE95_1osUJ0S`&8awBLjKfvonsZa$NqrK=?iokuYE_Zg>wi3tYWW`{5;LV
zus0(^9F>|^3bH45b@~HZ%xn2b^V9aT!gLc)x+?fRU1+zJdzTKx9WtqI#d31n3Z8ma
zLS8NpePi*Pm@>fgQd(Z3&`!|%IpW}xQYh33*p!!~WQ2Q!Y;|-L*|jaTYP0JTt)+c?
z7eWkiQo|6~t!5buA~9+nJIby5CnGqiahD}m)HicTdS1WGM@_2IQ7gO73HFns_9H!*
zh#~0_o9n**itTysbkr7hR2}fd7kY8~d`biu-2>H<#=uO!1u<Jc5!5MTc|qcQRLYQ~
zRcqox7=8|;Il>|A{@t9rD@1X3vu=s7d$y~^@dL64pZ$gXeHSM|;$P8%ddAupb_a?Y
zw$zjBW>@cO1-?Mp;Ff?(90XzgR)KX9O|_$USogbRSis<rpp(kqQ418=r4DNoo2}T`
z$vq<>I{#dQ{BP^f3ZVMu@WW>gW}9$#<|7G@Rq_njBO*uYT%i2koem07Mh*U?oiG$J
z5S?KKJ7oHHtv&}<6BIjEc3hJL{F;t$^V;g`_aG4-7`n=`V7O$QERuAHRbpxa7C{<>
z6^*bVvEV+Lmg>)I2F=cnXocny1#NY=HzAUM&2mdVdd>2Tt&nF`B!D1wLvlUEOo&l=
zv=Eh%Fz%q{*F;G6(D~$E9LMZ;k`;)Hq()7h>#Xt&!%$JKToNq`iUhz>M_-P(?CMCV
zs~h_Rb%Q5jg32XbdySv+fc3J(gfui#sObw7plVo!SM4CpaH@RS5X0*QWKH4XV1b%H
zOLJ&}lMD@yMi&f?$Ov#BEj6Ex(%274EY`R!yAg|lcCS@`WdS5x(^ebRYnjhsnhP8W
z;J?Kau3e%o5pm=TXKo2?%c@RnGx`mwLMvok#(7|#o1a>Xc`ox_?iG)Kf#>PJ@so5_
zMkuQ8bhZ)FxcRLamqzSsO!Jd=))Vwc@b^{W!C~ms=^3~D!bXyVm%kfCVvH$m7z^Cr
z&zGFfZ8c86)(+;uAWxy^4K1I3c`}6&XBgfM7|zePYwU7q8`Yd&Z6vG>A}74uW2BJp
zvlT<zd7W%CRFGLW+&{!r&nBjyzU;w|7?jlU-H|?njU~+M9^9CD#kw@LQz*jXqa=Jj
zV^On}+)YV7-zt8Q>E@PBvb$%n0)bg5IiwOV@9C=09==zdvb4lKoM`M+y<t1<o)4T3
zWP}$9UvuANhakJ_v{~mOyKX4Q>PptTOzynlPb${jtk-l?_s*a-N&@zIdP`mXkIU-R
zIfy`g?903>E!?0?uOlcFV|LfS)y9{(9VX)M2RgIxZUTC$vZp<A1J|7I)^J#!If#zi
z0kjYx_|f<YvJBQ0lbeL?Mb?LmsiOGU313zAM?k1ywTR&B*^~B9p5AALWTSt56v-Yi
z&(Mp;WvPbZbslTr#J(jq%pI#WAdv}hbF=1~Egx;%GG0&C)$nuVm*#kIeSGuGRwJTB
zSPXS+4spr4g%oIYt~H~$Bc2R&<5vnEIyACIeXs#<KvqJa8=Pot9@Z3+BsM=sxIU**
zll(z`7+zNv%f;k;5Zf#&zA_YB_lo#PYff4}Dv+_Q(USV)q5?ipO&}CfdC{riw5Qmo
zjeKpD0|~E7WlV3c8{`Rb(bp?L_PGZustsxrJqd|v6o%%OVLH8Fn2d=1;^SN#{-W6G
zHuIzlGgB-(SIH3c*{G5EzX6CZQ0Q4Wz-#Bpp8#h&5)yUd(I{R{z|b^i(%ubbxM#4d
zX7ZTaHbA$x&q{gnuG6)L{;s<)V9oQjT(7m=lxmqDOXAKC09{-pp*jzg9iGZ!m%AIX
zG_XoWSzg)qh~aorkh4S!ybfkXR%KfuP>Ko6A;FywXtJ*-#V2gERhRZZGV2HR^Fq-a
zgT#iLS5}*CI|N&fX^oAfhwi_XC{bg4hkMN`d>j-g+4tT%5ynk-`pUi_mVzFFk=U*m
z_GGKu%~jZTaeuob7><W5OPga@6h~d7UynC^H~VfTED9Wb78-KgR*a_Jf=TATh3W1j
z&8C@m{4?~w$2nc|yqP*<64Sxsbv0m7z6g{ft0<^TdfYL&MVCkKm^SZn^RC!e!C2?Z
z|D}|lDdDW~KfY)X0>O^7jKUr=dHp;`cVFXJ27gYQ>0o|rjyx3mut)hI)`G<)_2+_A
zn#?btLmtPcdU>BDeHFhkT=(f;#AzL!RBB&d)p$BcDU_@~c@uF?#Tga#-{|Mxf9g3l
zG;#2@Z?WoqXCx>K@%&R;u21CKe&OnQ(WsYwdtRBK8D#q!mrf&k;vqOqC-ax_IO05#
z#vks-ihKj_jeqxP6G4i40-n*ht*%U5(`RiaUHpb{CIC*EsCDB3wuF&nEEJg`T4or=
zm%zb*=s-E#9etAMNT!f$PRJzcJ%%{L?K{Hd1ysQpA1EeyffJ+(VCV3h;8$=o!GY1K
zw>L-aWBFwmW~K51hX$Go9ieJt5nLU?hD%9ok^3iKo%KMqLEJldxj?3_#BR(f!Xt1w
z7}KAihqTrBQb~GobyQ^u)7#Y}zwSe802_37yG-i<=1&irI#`l+&3#<_gvBGF3#EXK
zv4TnV5ee!u_dI+ce=THnYf$m3m$J!=j!<413M}lij!dENl`l4@`$FZ@(nx}*SBr-f
z<#GSXm?$<SIhD|$BJq>K_&jUtJ7Ynpl0IDWutchK+f-pEzgD7Oh<c;xKt(j2aR*fF
zzd9zC8*-?i8STU%__~D~&)bkRtCBSILRyKZrY;zGck7$=hb`GW+YXQO`#CaRamXQv
z?}v$as7unEjjqAFCs+o{93ZH%pv2tAGbRf4wS<_+8^*NK>0aYsxB%!V^qT?B=Es`H
zbIa`@tlmUYnQC=B2SEeIlg$GhV`|V*xSC?Pi#2O+@CU|tjLg?vtO_6DLkpE&e$%5*
z3zlKQ6Z_&Gh*ef4^69XJ!rs!_wz^07xV?#EY_7FB7~5(5cvxw=f8kctwLJf~!Y`>b
z8X`UH$&`?A>`;cbaWX>avZ$4BpJn5m9n$l;b8w6XUp_p={wBhXGshM>-n)}o@hrvy
z0LNa!3gCdt^9SMPvlu<UsgqlDAR2WWNa<L#+rI|nvpU^ORnYx<M{Qe1DWwU<6sj$?
z>!u!Ky~l`n755Uml)HDHddYMK>emASOy}1>7y(Cz>Z#QH1-S)YCZA#<m{)Mg8~yqh
zWai$q5ojbLf<06RQnb9EaF&yi#1Jsu5)i60yG(WNJ1X*fi6$|QY!9nyajA1d$F~Qg
zibH8|7!Z<37@Z+W5cH;m`j+LQM3M#;1`=B1KD<apYG0Jl%!70*Hj6vXYNRZ-7Tc|J
z%D(CY`i`A_Tq*G8K&@Oe)c~Npv|Ya9xhR5X&&~%#{KVC>#R^70r>)j!U`=h?@F^T)
zr)H!C)ab;=jMERPUSYA7nQh%g9vBy&sf{lvWxEE+eAP`3WaRr9*y{bpIg+zVaF_`z
z^IjfViUsBy@rhN-6611M^t+$9`%Bfr(|DR|R_Klp)2;;R)8>6T(-k!q<%)DQ2DaGM
z`F>g-wq>O?)0lIHeHk-(sm=7Jn`@7c%qviHHqvxuCj8l6q}YA3UoC7cr)6D8+f}wg
zKC-4c(}@&QwzxuZAgMGG?YH>s-;p^xtzqTE@62o1CKDi`5*y_vv}=ea_X}F;O#e=;
zJyI&_kt2GVBsuvG0;^EqieeUz&S#0QiSn6VrG~WHTHdK<q?vj)5ziR??)U0$weoDU
zAgKN0^ud+#H|V)DW^VQ5aNJepB5y`1QMt5yAa2_Z*{=~EEo-xMe;>1_8jO_#8O{A3
zDlC{u2i9t8D~G@k4C062z9YLgDk(@W9O@%^B!PJ^R7j>LW<VbY(kyDG)c%->$aarV
zb#<alz5Go5MZ_0!RiIwHZ}1REZOX7e>D^`}kD9AMQL&1Jl6$y&Vfo~K?sOc@N|;Jh
zPvbzLns*rCg^Tt|Jfq_PPoaz~N*ax*c?=JS3E?AmM8aVQ1rsER?SQ&7H|AUKWyhG5
z^uV>>C=K9?op$q$Ho70fdtrsC!$lVhe)kO?$bSZ*0t<ELnr}-p%}LiAR^QF2>};bk
z&>@nn%ri^DMukJyfyyA(M9FbvWYW-I?ui@sd32Ny$3D--SQH(()90Pn90*oUC%oBN
zDnrIC5;A%ZiI||7cQNS{ORtQnR(pJtt%=gKM%ajr1&L}ty{MvY)f*1{5{-;uKHBlF
zZthQLtgbt<eGYpN)oO6@gZFy*BX_BPLU@8c<!US`=;ghyG1e^6Q&DHC*+f#;oCbz~
zY0kSy;u=VBH0C1vDrb*SVuIe1Zlzl+Pd~5H5EIYJKA@*b<ND8}UoY*`IGUt97~&E1
z9oMv5`ltk}Dz!ta%=dSsl;fPLeJ(S*$P|Kl<}+BHX6%hIRYmXQ{p_n+eDvLOCFO3;
z1q+8mozO=_s0BTbuL|Q`*ZidL-NVi=CGSn`?5OyUECd+4*!*UW@tJ~$<2{<p1^w^v
zlVns>0yS+@CLI&8jmkeahDuDJm+0MB>ij5_1(EMs1*NX$6YihmQO^-rpsR%gs&^;=
zwLTi7YFc6A-vHNI)&VKBjQ)#!@skp0rUR7f9X6|KL`cre+g@!qo*)alDGi5a76nO%
zMhA7`zL@Fi2Gy10t?G{egVo}d5++=4y|o6eL0>5GdIxj?SG@oVhxFJJGH;~q=-1jc
zZ1qv1i_w$`ILlTZO8?}|NzzV}%$^l??>q&%L`(5eSzY`L{JYj2aH|sc7lEi{pyoH@
zzpCVuRHwu}%D~&u|IXv)y3d<*dJ5z0h*CopQNv9?K3OIAMbTU~fqI<P!+dnYMlrjO
zIv5Q+)Jvc(pi!m_9Q7u<ySDL(crEr&I&p$C7pe!J2Vsh1SE@NymEwrr*s!lHmKP}d
zbRJ~j1)j`5h++h{=Y>RAJ#t|cN^;`U49uat%H86<^;pa{Y<>`Xru4%^TXB7}=m>2U
zY)pQs+6g!XFzhM^mfzl4!Pe8QsC2I&XRxiDw?12i_qSYEy-r{q0cDo~s|*gD5=xEt
zUg{GZywrb34R6MbyR@{-H%Ery<85>Uoa*(8tBG#(p+f|n1alwTa_EpjrH-X(_!Nxc
zcS;360dOv5&o<GsC!Ghm3k(={WLL~&vP7uA)*>&QHdzER$V9?!9?%Z}RP3?<`W$Wd
z`g6)i+!v43y^LFA((Ig)*uJA){Td~ZLz53H!KS;VvB|sW>dRmp-->ZFa+I3OtjgwT
zz3t;yYov1iYVQcuG#~EN$S;=^n6}$G>|$BRHF-BE?7fknaaXnT*b@M%<D-#traPKJ
z@%HAb%(zhjMZeq|l(Q7W#f(fZdXD>GN!66-0Jk}@F$Z1Vw3bF%OE)_T9mMvnp!j^+
zVm|v`z5WGI^3SllW}LqQg+0flKT3a3$b5QKj1sc!Tfv-m9CUl7w+z=sWN}MgdeX#Q
zOOp)fJp~sbJ1g@5z+wScj?n>wEC{But777K`MC9Cvx!_!*BM}W)Uw8k;e}F~Q<U;x
zslZ$}q2dmiWS8>mv=GNlNYXvwKK&9uata8tvLA$=PZdFEq8ka35E99xZwhf6%*Jbt
zR;i-i#Nv0a(cmw7(G_={X)t1ZUv9qlEc6x3K$dw4j~)`^US@m15+L8P;Z*J3sTB9C
zbW?I!8_NiTUbGEm*gtFc8LZ<<UoRmw7Ey7W&1SnMm0RJphH!0uaw<-YmIxDBigLxL
zIeTVPW;=;+kd~EjnnZsyG-Toeh17%SD!wZn)`P-qohmM1=URT{;*->)7hNN!FbXTY
zgT8c0cgk}n*W>e4qw99d3GF81)Ea3gd1t`PL=@%hA4?Q|>_o%%0*gK5z#XbI&4fxw
zLzN?R?0ao!$WflMf7Wtq=zI_~NzY?iJzOq?JA{&kc_1VJgc5@`$p0bO%6Jb#bdO}+
zDc@v9>q4W181kAjzK&LQDN$p$2%%8a>LH_^wGHyy7j_|b)g(P-4JDSNGw~#;g?>-p
zZ=#K;ZlZFB@eq%bnCF=|>4KV6d{f8<_9oS?EE<iJxwySh+uUri^g)v`guKkdwh$ft
zs9}$APU-S%L!*(o(ld8yx3u_c$~SCW9!t|+ZcInVYh`D8K7#n7zpPEj7!gz-+@xm9
zYBHk4mBt%7hKZEQQ2T*58)WgsXX47OCCTG4OX<q(r18`Nv4!jL{4X{v-^vvW%`DuJ
z&URJx@DN$e3BhlmN6c^Uo(8{cGm4CgoN{m~@>0_SY{J0Q&)v0N)uX93^c#J;<S6Du
z@7FGR!JB=AgsNqBlJV1Z9bKO3wY{au+yT$j2Ax4ln)r(3(NPv9yc;+db{<*^D-m$w
zvJ|?1cG?Ck4kh&R<QOcH+lzRJx}~WCpgS8!h1ScPlTAfmG1gJFvwGeV`FX|#1<0gd
zZ%i3JA8qrHF;VO@g(AT44;fl3Dbrg(@#1uvF@A?Jc$@1p;2{Wpv6O7q8Uv+f$}u11
zvx)pucS#H)b<b~AbF6OobS>85;W-JKQ$t)rNFf!13Ktjp>7gh&M0qm$j8r;*4$0e_
zbIx8N`hE}!s2BDpu(j$tBP3X4RJx-4)#=IODo&oWA$G~_2A$<u_w|(UHH%M(V{?nV
znGe#Xv0mQZoNx&wC!_>t4jqRs&ikrrYp!miPz=hFlvuHM9tdJ8i(W$vv9#59DVJgV
zTzZ_1RDu(Dw4U8bZB%9h^iFM+f-iFkW%s^f7|jCdip^z|NNv&D8{zue=#?5Jjakv8
z^+o9Zkb-C{7)*1Y17qfyQxC=5(o`lhe6PPZFf8oF*1K@~qODq$)q$QI`)SYVd-*W3
z6YrBlFEFMKYxM$#5jTiOO^J$gps9TU<3*gGrs>;t%)a`zCT!eNZw2biM%HQRfpZ^p
zpA*nC@%vyxi9N^=gFqf;k&8@o=OWwuu;y8^$ks;_dv--ePc@f~<_cqj*v^XWJkcA(
zT6Yt+l*8g>lmF9m(x^1*6rV6Bh8-Ox@&&4re?O*;2Fk79BN#t(u9Q0o99?f>XF~_w
zgN4j}Z)%9S9e0(W&Txj-sU`25%faeaZMT#_^s^4vMlVi+=psl7Rsjy^<pe7x=Yumk
z8UJN~gjDye#Z7CNX}%$VTC2_7fApT6j@3zI7V}ZG*|j|3A>#nd`l(p1o_dqHYmu_b
z5CSi862<n4Id`#|(X0K@lR)*_g~VrRF^S=fe`Y_@5c&`gNoNW9yvr2Wi})d(PZ(6#
zw>Q2aQI~*%Z+m=SKB}C-zt6B2%eI`r0?b+!&L&v4>?k2NBD_TLE!2-dOK3BpgtYm2
z)#JU+GEAu1Qwc$hA?bndE7z_@ZuS?xv_omch_rMVyM|X|pF`bhC84TUZ4b#UL^euj
zs@HzcL)}3O`>;3f;xH7TLcXEdbg}bGJfVn=Y1fe06i_x;2gIFQ@@<>2s8CC9McD_1
z3Cg=(3U!H@VdHOsDfLD<kw&MC3Jz}{4eK$bdkma6pQlF_=_{;3%Z;xrP0_Rb)-S-R
zkvPYHps3jX6Gg?!{J$wGhW}40m4Sot|4dP}|4XG>p|MJn<8Q;jik|`iDlTm6EHDYd
zF!e)k;}kj*fiDo?Z%evKM))U_)8OMDvt4kU`pka*zgntkuFtjAYR_x0*-Aa*`Pt({
z;9jdf2|M&SRX-Rq*qGHt79_wQ{(N{mc=GVc3Cq3k;K1Jz2aTBi8*njp>90QYhZ}t?
zVo0TWfqz8_dvyM3AV_$4K=ANzsPIUj0004e`71lZEgkH9{ljx`K=X$`OMCS6fN3hY
zHy4LD=C}UB!|$I6{RX2D@bE}T$L}_<@h*S5xHi`N0Ww*|!RnzX3v*zQ^WoZ@fQfv5
zjst}j{zN#VoSwcuJRCGR2;2>9U*kDF0CZy~zyk2l176)gfd27Kfu3u0`17lRiAPJ-
z-#>tGFIE5-hTmhoH2{DN;L;2<1W%3xhYo%P%P#_UR%Hp$tTRy0XB@)^jN7jl1_Zx{
z^ftP!x6+5rG3bl6xwbx_qiv(l8%ys6@D~9R)SSxVj-V%j8vqOa3~e4X(!c1+hT8_a
zfVGu;@OH@tK(etD082T<Z#k^q$<0}W@C`n8PLJi`*D;9GR0Lfd_vQdFlmMZ<??K7l
z8Ems)W;^V+?u2t_PnW*0k0hK38u(oWbZjVP7yjhz<d3S#TU1b>{cq5EPy&EKetLL#
zJpjNJD1g@{$K5Za{@x+vPt6xhktO(@Q%DyeemP_RpICVM3!qgvgTsG%zipoX(fd*1
z0|0Pr5d3M?_Yy<D@#`+qr$35|zW7)DX6wfe$K?A1aJPDyM+UVN9I!`UKI?bZr!_@a
zMpsu`J@BnI{!4y*w6_Dmrz^k!P?ru4|KEQT5FP~u6!>d1E;i;+&*Lkth7uYOEc{yn
z%|+I)?CO0L+WxBrKil8$b0M&Z6b*XsZ8o3p4<yjFydB{0?BUPm(XZWvpT~P2?8mQZ
z@zI}<pB1Msf-k=(R@&Hqf<N@~@n`<RR(_<|4Un~8#h3v<RW&4N$0z%Dy&B2|&=#0R
zdaotnM@a4?Al|PaF2ZE&o&kIuvErPeyo&cAte*|Gaw6zzATjozP6K&W{=MHbc`4J=
z=uQ#%^7*eCz#`YPT?Z5cSNpj-6C`BRKWpt>n{RuCo-v8Re|!PUYk&iK{NR}U^WZ?m
z3Z((cF@XT!+qrJ)Mk2w1-c0-i`bZQM0Qye&7KjJ{`hNHl;Uxv%!@RyNu6FAI<bnI`
ztPX$l{P7{}+Zup`2I7BDAihC<H*r<a`6Az~ihc)xK>)s9`fil!xz?tESTKLd<&DWN
z*neUQ`O(DE!4CKu0KRI=>Fwb!TtAM1p5j1Ls;HKL?)jd?0DRr(fgxW+-PZ{F{T|*{
zDK<UYyNa~Gb-|ROep3B*?ey!})?nQnU~S0|Zk|ZHzY9fEXQJG!psieLm`Dej=`_*D
zS$`L11%SAuJT_!zFqK7k5^I~}Q8`#n<wjlX8E(FhbY#aQGbM2mWg6^+Zo=98&PK<%
zyy%I*Wv%!KGI5&(9Mj7u`HgLSN3(g>-ETMro3+e8pHBy~WdF8itkL#^$a{-UC5wi6
zlV)vpMBOhXN0`8HSbqe+aa449goXuF?*D}uQ5roxb-Z?9=JRHq$}@wdqk}tndM$y%
zK`Vb%j#3s$oR9ykWH^o`hMAKmJ}GA*m|Hsvo;{jWL`t`a)Fxb4rLPfDt2KrE*hB;i
z=El5gITE^S1`@;iH%uOMQta5K`qygAS_kEOgDB5RGQ@VbT+ldDuM#jf8^Q=%7M$zO
z<w}kQsP3F8$;)KAOI&bLf{BKUaaap55KPS@tDvjSGa-Ql*DRJ4z=QGjBq1vVGI7QU
zK=*4;l?I8QZ%jNdH?(BnsPTyA-rN8Dshn*XJo06{dXjq7|Fm~oLdZV$4>05;9w;WV
zYI7rNkWhKM;)y^=D+F&;^}*dS8OF%N;zKG<OY;&dt5MPzloqu%NaWZN4=3OCH{Rhl
z!lDtn%&l+_w!VXAAjo(Wz!<uOOCc<xD-p~pQN#lQ0!kx|n?rGaJ&3SkyL0aH+iH=n
zci}rg``e&Ad&%rnPdjg@Cd4akZ@nec{zGWSiQ`Dj*_KG=W`%AnqKf%O4EJz#@&0<f
z7&^p_wBBjXWG*ea>O2>^T%`~S+w^n#gkeYa>dCpL)YhYdNjrOkov;o|6$M8hh9*FT
zc^IN~oc2j!2tEC2!@MDv<*?NyMi5iRcLy>xOM+_~+OFwqF~{lLL+O^J1Q3tdKqnjy
zVWVsT+7$7=JrKtVsfhpf6Q0~}+DM>m4?7Ze0$}lXJ6N*HFQ0mi#mr*^TT0Qb;>y!`
zMW|1DMN?|Fj&;w<Emi}$^66|sN$l*9%ooWfK>$y~UF}+6@1`}R0V_wa+!RW+`Fv<O
zF=@;+5peNseI5`AXP77kYXF970JXa^m9bs260a5@<&_!ceGb-3tqUy14=a1`;A%nr
zk~c^<)o{Y$-q%#>3wq=nMQcJ7^r?42Zh}kzlNj9-7O)N@#nGk@MNJ0SUQ7Vuh>jXy
zSVU?Yn|=Q&H(tXXoy|CqQ_RWj*H@>XLpgIN)`}$aJ%55xyzZe%q(kb6^xkVU(f---
zy5Rg&y3*aj&U8oHM#*oLsG~)223^Y7HO6-m{VsmrT|V{*lZGH`fXO@K@U-v#yUW%;
zW-e8c0-34h@hIUnN0Reu(84$|vccNIZMjnzXA!I@Wn2sJeZ}B`p2%J2Bhk$2MmN#f
zO=rR<A~{>LA|Wo$WAKwrl9DrzUhT_Vdb%)kVYBbDf-J97-kD5HZQxH@sJ-Y&B$=DP
zBXL;yP)71Y1`Tj0O_gVXQG|kS+IdU5h$^%qj^e$`Nj(=D<YTX-el!l``XV0dl0r?;
zTSZ)241qb(9o&d?)#l^t`z!bEuw43Sn>GzSt5v<_>ExAZWv~col4R{!J)fEvOC$18
z2K(gVtm5pIzIq{rk7gTHPwLeK3mWZ(1iKtbOEHl@_?kSaWGcmyBw{N!qGSxiTBR4T
zMGWxV#~G?vNQ=2s=qi#jpokf-Toa<`zj?q2UAOeYqmPKIBjEc58jFQQ;ojZ7vSU*}
z9g~v|t&6c4?cRgUuOY1<{2;C0K~tly5zC$DNlSDH&SBA!CVEXdmKXZ;f6ufp!)7)V
zl|S217@b1>ix=YLb4_sC4;E1>XJ^oFnL5Bl9Jc7E`REO<bM#na=)5yL0PUb!vdt_E
zqc%A?TJb%AWny-%1)J+b!uk&7)b-kqrVJU^E?`;-*ay_BxYe_(wTht6=gel(!haJz
zuMHwf@q(zRyt63?$wV5J_aL+;nAiF%CZg+d%*|rcIq8U+_C_%eVIc3Kaxd@_$Fn33
zeI@gb#lwA-OwE{XAh9$z%JbftFV$r8RZt^c&-x?Z6n>t`(fn5!`$|>BF*-X`ES?OU
zyz#3iOQ|Oho}!+p;dn=W8fX~tGPC#jsT1FnxHg(LD}kkMFU@n*19|%qmCW)aR-eCi
zx_~<$QQZ<AU)&HyTud$WIzQNW(9nmo?|sB`l~~J(-`geUD50nA01|V|UqBp5hvS|G
zX(-d0Y<yCVOENTsz6T~nbfxty7e|D1W023!y$#;CmI!~E4tY-HN1cYjp!6d7uTprl
zbey|sV%~xR&A*&|09Nc~feKN(&JIv{75BTMg_W~(7GF@!;dwLNy*l5fY>$c!0A+i_
z6)>`D3KXAuRlX0;C5&jzM%=91v~S)?xRmmidK+2`yRDtI8L6h{+p(;4GdQjBD-c#-
za6T<&gs+b@FckZo%j3a`d$g?t9m1lCrkJS<U$-<S+XtSmdlv$a&^r7(o9kv=q<Ge<
zJpXN(Ov$K0(9|OW&>L6TylD4r%pDGe^$LDy4el}qdK$x+JGE+%RrcR`>Bhd24*A@7
z$as!mrOBs^u4ELf>%D${rU^|GO@D8={gyF+j%ghU03S6$vQaw9iS&M#l`i$OG&G0-
z`rPC80J!K@3kC4?QYrUsv?1GmU13H|qvo&Lmer3JB+JdqIfG}?x7>z#W-awDBs13I
zdcO9>iz7($(~F`y$y!(+;z!bcjdTtRZ<|HqG}`^7q5(OLeuZ=k@vQiVeO5=*HJQFA
zPmtCFTFND}&pHQKGDA$)r;}6NaKz2cuG=SMdO0!$!r2V9oBRm`$wI4sNYJ%W^`I$k
zgdZJ$;6<k03kdruB9$<2$G)r^PwA6E#_H{{6!qqfh_BZ~Bry-Y8o$f?X8Wqn7|zO)
z;sjQj;&VZX(gJ4Bt(a{|ie4Uum?Ov{!l3N5n&yVRZ(D6JgeU8O?_UXlw2(d-32#&T
zG1^}ihYvos$xK!9PKI(SVP2PK$GUSKwBB#-^<#AGHzAj%dfr2Ki!xzU=Gd25hL9mS
z?3VUJ2IMy5xYkI>WVlot`n$aVztDmmJPHXMv?B-M76cS_A@^q&zo?HU@tU>_DYqY6
z%rc`-JVES4vN|r2KXcVCDUJQA&8s1OzsP&#wxw2m-!rUW)BH&=49Iu2t)HUqZ?-kP
zVT#iG7YFiI(P)8><g*+Ap;RAucux#3gIgJ^Ry$>4p9Gq-YVBq}a+zR_jEEBHgMHql
z2M-v|%SLsbd_Sf2^yB;DRJ}nF<!7nvYs=ANTr^=T$@>d}{`gc(r06kKZ@WcVMDDgw
zFs*z~Tww7E)x4CN3>3$==4*jhTmW5Ndk?#rod`(hc{~IOhSG<CMcxV}Y3o#_v{Op<
zt=1a5<91rj`i~eC(sv6cT1bbhqpPuO=SIA}&RmZIU27a@8&P3G!$&umA@dkYShGsV
zB5pzZBHQ$rdu1o#87EzNSuS<B3Q8|^F{DOguKS=(%Zi#{@Y?SL;bhC~El*=Mm&HFB
zlTuqm*CAG3<Kp_YDBDx_FwpTLii!8Wtx>N{KgD<t(_bv`VbNUwS%EVsy3-j;Hn@t-
zk`?cNe{Swunr)|Ey3MdPu<^^mt)>g7Zm=VXo&{oUOYE@6XusAm3G%RX-;sFdQUQld
z8iNaoH1*3HmgAJJvK80MCxbosoTyWl6k}Ur1f&)hx;U<G88N(LSo8|G$1%6)1a}G4
z3uHUh){cW)D%^W!WNEpyAmFQsbu}&v3YJ`Pq3CrTXTZz>ZN!eCKy6keWtZ6JZHWg}
z_V(C6b!pp(6jwj%7C-OzEv>rQhw7N3%UbtpEI}!@!m+g@jTH^Qsq59r@Lco9nM7Af
zr}x0W(*zUz{?R>#n9cUp%N;ux{nYtLy9psViI}|QuHoBnV_VYYDdN0_eBf0xea{d>
z(`{xNK#SjHjAhGQP?mglC{(hbW*V+7Wu;<Jj_ka^366mL=3NyR<G@4xZADYHi{2?x
zh0PcqF4O1*cAdsy<k7UQq&@GLti#Qshv_U1?1Pz%7NUTN@ne+38n5#1aKAt6VOq96
z(2>ZC!lr=(%tF4(DO`x<?LGJrf-Y9|QLtlHzJqaSYO1ajK_*K%0a1`I@*OzME47HK
z2i=%6Q&-QpZ?Z1YPMyU7MO?4DHl7P#si<4e4E-We8~ub_oaMqOGk>)b_q(A^bt+p!
zPz7O=JOTTt@46x&W`gMp4ewyoa4}3iYxR5U9oMZ+snPapcV#<dAQ06%U(+Is@>E=l
zh#xm<@zy&R3HjEisEGo9g4II*GsMKlMyaI~J318=*WSy0jg9F<`;5@(UpK0o?*j~+
z_qW{I8R;*Tdrh0RIx^^?+Z>7fRiN$%1*rlw+B5evq(-sEB=Npc0lns?Yre=y_B&@2
zJ6DrB_f}15yOS}P#Y|r{scll9Q8uX)x1%G;(*gDNGp};nhe_kj=~Oybt5pf|S1s8Y
z3>n=yOyV|YN-6GkLuXD3iNr<Vb>eyN=#R!;;m=mww3?t}D-{~m-(&ZoEd^)iz(!xT
z(`7bS+6gSSBz+*~H3r2*Th^KPbUd+6>A8?lCO2b1#|_1vyv|l#`~63MqoR$>=O8@}
z90v%cE?Og4zId31i)}~9_0+1p+H4m;&+>Lp^VsYU{kj_2U+GtrNs>t{3daq(f=7-8
zE_@6EE?GmU0>=&J!5obb4Q5r^X#kWD+>NO^kq*R%d^IZ!x3Dc8E^|xZu^iinWO4I|
zuEJ|Vc<8YuACs3ymv=IYn%Zo)O7_`$Tc_$9vjk-Ol^MroZBsOsJqyd$1J<Pwr1od8
zMiy1LyvaIS2#W$1%XET^ZfIxjFGLE2>l!2$G|mpW4i7!$lw&MjnL3pZTla#Ytx{fG
zjB%ZLUYiXFRFWi6Rr})HefGe!vqt6%lB=AC;hhA>^vVa)zXCOSethxuxg~2q)Laoo
zgu1^U$C(w7K{fo2L!Ma^{(4PnYyP<<=gO!s1aLP}s3i-p7L<@tqoU0VwWDjiO6N33
zoK)44iZ%t?K;MoE>Sj@9?-LhLoUkqdn$IWGZpvXbu7yFz4zbYWK5C@|udTNdGsPu;
zb;9zs_=J6^s34~*8rpWNX`0l*&bEgLA%}n9pNg5&jdi5vFckHj0AJ;iP&^TYyOsQ|
zI$9T9FpGPQuhNP*nh1$`@P4bMAJ-1T+8JViIU-KKgTW7?yGSn)-G+{D$hM;>7zm<H
z7d4bp;%AH)$)iE<_E`wXU+HI|X;}5bNF-!3{_Y%X6GN=_#lX12tqAkJBvi&8ZGbnV
zj9x)E6QA+F%#U}jo(QdD(S1ZJ)yd7}W#`fz!OU9)>tBT)UlKP(5GDwto8eCMB3`_}
z@nuW%7$uuO>>2-Pk~zF9@UhfvHW?kAeRO&2=QyWns$=;jA~uB+=)v6Gv!aeryo$MO
zXV~MzJl;tUCokEKvHe|5`KGsP&m7+H=2$FWZ3%y2)azf|(m%QpkEb732$=s6*T8Wj
zkx)=FS>Ch)onPY=5;kAC>}d4OyRO;aF}4osX7|{)GI9HK5vtZT+*e3_D}Q2V{mpy}
ztyRzNUots<m<|!^IOhU>Uslh|)vCglRUS*Mk5K^_Xo~<`Yx}m2a){#9iB>SW0KX0q
z-Y`cK>}^+@t%|)l&u~+eG3{?5&cXE_=`exYzOg|oC)C3Crb4jK`<F|OFr9<y;()ye
zvG=Zd(hxoZ&Coaz5`8xREax)u(Sz)oY>gN*K5&GlR03$aVl|tp{=WEG>Ad)p2MNj+
z%Y#hllxPGFB)b-9=ZP0_zde=GY_qxGIkD~lM&arpbqEK6Ce|b{EGwhwj4L;IC3jmv
zX%i*oMx=y_SpNjmGJ5?1kpq0JGZn?FVjjf{>%iLUOiM|A*^kapVG8Hv)GsfcAtXFc
zW1;?Fs%v$g0E&Io9Gp&uOc?7Ko%ik^`=YghoVZPlB1|w+L$2vn+!FB*Qi=3BS~YXP
zK3#YS<ckEY`^9pw-TZQjB!sUan^sba7N=fpZrc3RwO4Cgtjmzji0t8{6HY1KUqb4U
z<rFr&995ZGzr9L2Hq`qY^Gb7p&CQn*-?S97N-Kr>F+WTaMy8B*^2ch(*g{j~nr!Pm
z!>l9*q_@R4&WkVT_`1@?9H0(8gx1e)KGhV5on6gyf4hHf##si4ljuHoh_qD1y5U-3
zwI{32h6_cND=<+B&B)?=%Pssh{%QM|<phbNWuDpcz0G8UF{C;(qOyQW4<ddT1}azG
zAfZXlT<gt-t<*D?B_^3XG+>Z#v?E8Q4LD{PS7-rnA`{?m5?W78nc!IYNMv5kyX<ko
zFwf5v$#b3@9O8-nBWUOpg(<Tk&I5gJR-)pP18skZNDhUnT@qeBs6iDmkHBHY6YeXY
zh%!<4Ym~AhNy?H0D{B$zP3$#S@?J}NgtVeqe#c$4G{@N^W>&%ol3{F>LKE9^Hk?$7
zt?$#qhNprV5Di+KLB|}b?(#bH>qSFmv;y!5V?`bKuyicFmaWWpy;~Ow(25}PCkL^1
zMu}}0f$O99`piW5x5j(Ic6`j5*9KqjwPq6RVRAdE$s0>uYGtJScv;2nn9+mRa<@If
zXzd$&^>6NR%MHq5a`%YEfP^7aiCgGuk6g|$G=k(b)w<g{IAF#Q8%h#hoX7{8r1~zw
z#n5ZFVNd&S!A8ILt}X230X=dw!en)}t@!1WbqgTX7TF8<`8f~O&7TFC;wSpY20}k0
z07F??Ka<IxZ&|Jd!#|qW-w`^0FWrUy9>s7wqy=&ov9eBgk&Cm2^+czE6t9sx6>oZz
zk=wJ$@}G~YiAoV_5X??^0bOpa=PoWk#RQ(^zP0zJX@O2tr4jWHn>cQQnXR9ltoQOF
zNi<Qr^#i6z%`^?3<MXWZSjF~&WAp)6bn%YFb;TC4;pvg8+I^0B-+Zncw!20=e$s{D
zqD1`R>T4#Fz~j)GXCz0;a`V2N*U4xse+{c(@6P|?lET5w;*y)|9Hr-FG6T+O^1_Dx
zRHEn6MeGDz;VZhZ@H64~<P6DlVoh4Dli=zt5%s!Jj90H~?#dnWx|^h(`8MC?%~qo0
z_#FTH{DSqJ)w*SU^6qHfdOe>}CoE+XS~+`#H2mImE9z>fy)Bi)a-U^fZ`?UxZKnrW
zV#@CoY83x*n&+i~A^5`4ch_|Rqg_qiOJi1wYJJx~Ee}ivp=bX}yMiw{6}xWXFlMDC
z*=y##(HgVjBjn|&Tm(ll?R3EfL_9b5Vd%Axk5C6(2iEwIQ1Qt-PdFO~v)jAFUZ%V?
zaIy}71tFUxZlO3Z&OYuTv~&d;&5IH#w}?p#Da*o#ix{4ZCv8f4vA|;MgHh`9=tFr=
zkk)#uLY1fU7?GqS8A7r7d@&jl0@3)^gHxfSuz;*|G4;_DHBf|ut6{KKe^IGuP3af=
z^lWZ@f*BA36IiDk^{eXJ3KIk`@q8~zBB+6Wh67+)k1`~&T4wN>pcL`Q0^EMQ^IjoX
z6RbeOes4QQ5vsbJL~Ah_b_K?vXFVCZL3-0?f~J<&`r_D<MUO8PW%ki-{EYM~O%hal
zLfCq|?r3d%FisPsr-^qQ=C74Pi^fFEeXKM%5qoAYa%yRqYkR}5bXr(*co@fFx)DMW
zO*dqeBv6B2vQEpzrHy(L3C;aW#0VzJq9cdR;krRaM}IjorXeQM$+ipbQ=xkJHx;I(
z%yY<Q&IeoR=OzCLu!hI+ErZKC#azgUPJ`<)M^!LVI;ilrHIhpTBYddu`AY`~E!FNx
zthh*#jFk7q+S!##Yk12>BzrfyzAL~3hAildBLJCEu@L~(@G-fX_id(NPSrEV1|(*(
zhyN;U352nwA!cW7Ci$MN036QkSS}JrHBXmvj#s|C)__Io%cDYU8#y)5zR5@R2h*%W
z4||?$v_07kHIsylQH#{L%bSq0c5Ajh_%v~O62*i=jYA(g-V1;F^JZ!T5UIJ&qz{A^
zgRmp5k(LE@ALErfgYvp{`zyGk{L>KO+N?9KdT@S7M3s{YsUbH`O#hE1EAyaY5)f)H
z;r%_JYvWl|2K8h?6|SxhNTG;n3+$zmGZt#2k+i`*p3k(c+}KzMTw#m~TDsqb;Pfz#
zD(x9B9-O0*e5XrKvwBcd;ze`)<v=P763yS8PfSsxDi<iATG5s%Sn#sJkDdtc;yS&P
zZ76Zac+S|q#Gkhm6<|#X?tURYa!X91tYFyce$^FqokNdYM-Vso0rrt#d6r>u`$G<M
zscc+tT(#2s4^sv7SJ(%|{)bN}&P)ecpsF63DdMx%{xqEFti*<6RddDHHQ6NUrt=);
z?foHNh4(`jzEt(g)H20ttN1Ug6pX_t=zZ&<NrE{n1CTC&g4569iI^+tW9|(%CyoQ{
z)2nUxAkgvZpkOGWr{a>C8(Bp-<3O#0;Ftv=&=}=NN*8VpX@V1v2&qR={;umU4^YsT
zmlf@(d|>a|fY>4T@+Dh1g_Vo-;_h0y70r%WamHgEo_9`X7)Auz43TWP^%{_p4>t5g
ztjR=Rtio+098;ISVqM9tlJBV`Aj}nUV2N>3fDrtx>LAv=r)e#sLsmv#s#i6pXh1sG
zX`M{HgG33E1%?go;9+kWOP$vO+#d+^C|zaOqJbwY3ondD0)wn#87wf2^(me-DKSvZ
z(#=v$brgGw>t+L|G#;PubE{~IO`M4{pAGf5=%!NS62jE|E!X*eMk6jb_AEkwNzR;X
z1iD00*!w;Bw5d!AKN9ysuB!AUy{9?sb;~bYQYQP;JR8K;*D^jyMot+6Ti`gHhF)6F
z(|0O0POZV#Lql(kWS0->=5%8x&}i0$$~T(^73IRVQN6Ig6Cz0Lh}M^8x`}9E0A-k&
zOujGL={(l#zks0c)B`y-K(iuFc9%T=`b#K}KTBqwH3!}bk(oti1yH`f5L*ZkITW;w
z)XwgmJ*-adfvp>jON?e|=0QpMo-CtckOhHEpSgY~KF#oh1T}Bc(T2JhZB(WR?wadI
z>8%MZ%*&9!gN0FRDK>Hdm(-`e_7qbyB8={5Obw4r3|vku+K!_hLGEh2^F{cztiq*<
zJ)&hKyXp;wc$>g1c;?_faaZj8h`B7B9<|^)i=5L#I^b+&ZGZ`c$ki;TW>F2rvEZQB
z$Ln?1qTBfoZb`<d(GLIBnsThIymT_#Atov6mdUFGZ-BX3@PpFk;{0n<*j9<?>cTVG
zy^QE|)_=W(6*p8+A>ggL9cfv73j+~zu!Ypk%-~Wq^PCFjTc3nJ(o-I9j__$?K>xUj
zB_>5M*I7MxtPK_Z3TteY6ksEP1%y*p5DJ@qrONHfMcv0#x#|RWTV~M}i<Xf%BQlL!
zF_L)+D8J#aM)&&jDmZxNaH1BdjR<m~_gZqC5#eRGl;S8oGx-7*u%i@GpnHklj#o@M
z3k_<)=a#Hx(R43*D$0K5p!&+oGjW3{W0miIkg!j7#;dPTVtd1@ACvn#JoQtdvTmCd
zqm(lN13C%c>QDVlMYFLX|9k$&G!>szsY&B8r>cjIu8>vP-q#tCPly^DlRGAr29&BM
zt?MWuGklT0HERdhH5aov5sQw>Iewi&=9nt6Pu{t=FqMd8ZSK*>%_;IGm~O}E!;}X8
z^j{VoMhEfRyY;Lrfq#(c<ETUn3G-AU#KQHmr=vIA12Z@LGDUU7%-3!huP!P*3vWx!
z5p~~Bph9WSAb^VdIar`dN4r;+W&bZVU=Mcv;TMIpUh@O@hNP-RP4mA5a%SyAJ_lwp
zORv<M$9b)PkJCZfwX3|=6)DqtcO^meg+bYnO$&Zf-!4(eex%q%!-DXOPBPzx*pY@f
zZBxp(j#5`3=<9(lnIQoB4US%$Vtn<BgXZD=ic55Gqv#@Xt5n9S-b0dInonD4FTA`Y
z$e?00GD#&*PrnhTOq8%Or;mj}6=>>J!Pk)x=$g`7S^$a^&?IY;j@&5Yf})(Gq>NrG
z^*n81#>B#`Y15boT@JJr0d0=>kx`hXd)Xs@d_XCZ+_Ax_u<3>s-o&wdUFaxVY9(%m
zqMZsD-0X>%(e5=xdxAq`^S?wnjXZ-C4t!W0Sllb@A{z*KBbHISKKk;L82&taZe^Hz
zyX5UT))<H41{6$@klCdaYhB8F$VQEW`?L`yN;bj{zGtxTWj=8X0AtIwz+Vo%q93nd
ze-eDGzMQ8dt)zuWmGA=|R!ji$R~N&r;ncv{Q`O(63XNMlIh5pGWF75`?i-HbNX5e~
znQd%u`_086z*}eSn+=Dj2-bBJ%EvA$+25mL3g{u<5LA9x_|GGtSZU;G{q4E3AN);l
zHUa1T>5TGLqi*`wl_KWIa*6YjH5TeR9RyF5S@pYaZ?y$~rKa@Vw*wkkwgx2(mD(!p
zyy8mY7nn!uz_FnK@avb?4UqlXT{lYu;=#_L9<I)g%QLx5{Luy^&%>?TT@kYJR}`#p
znLgVF&^H>zULU@hmqJP_uCi#wsI^_GjavbRO>!iPQt=4R=nvN^zuHeCnURwXq~Y<r
z!`SxrfOYqX!G|VjBdrNQS#DLnhgj6omGV?J6cZ^O9X7x^0PhTp<TCDs$zCRV1OzvZ
zxk4UlmS?f(Ueub!?+vm6kX%h0C#r`gH$~4uhz9Kk2vJcji806wMjxe*L}9=TZ{w_)
z1;V-q8RkpjbXn9<zid7oL8q9$b|`CNw(DmT-Zq7!cV$Co@=md`<K{;IGKt?-9eSjg
zrpTg5xZ4k-CyC1Rnz(k+eJwvcZrzE5W2Qhh8`pp;rp;D1fIPWZcN+RQGPg7(CLV20
zf=*N0U$7aY70UmB7jgV2yoia7;eWJDjQA`Z^vwVB`rnc!R(59g|1)0X{0}d>eWBeZ
zk0}lUL>#1V78iFT!hszE5Sr?zcV<&WTp&>dg+W}{#wQSWV+#a_{2FCH{qj5AX}Qs9
zT4CG$yfV9WyM6V^jf)YHN)y|FG=Nay?^oc)*X3pKukNf(0QZlR^G}P5%L|W@feUr?
z{|Sj$dnyn|LkFaN=>u;RCr1w%IqD~8mc#i2R^HADzz+xzUym*>932POKRfz*rY8)n
z2#pWs8l>@O$>R?tIQvpaxD4w4QAF_iTH)O0*Ck5-pM79|TpmP^H!d9fYjBZ4e0_F)
zoan|7&O*0fe;j~LaB%3zhc7V*SO+cS2?=O$XE(P~5U$QU!g$uiY`8x+^vN{*U>ATQ
zodPugeW0M`{x}7Gq=zJiKxgZN+W+i5;8w|8=nx?Jf56azA>9Pb+6Qa$#Q{8>{A(LA
z`4=G|e}z;(h1|j4yjTI?>CSeKe@uTVfrEZ!L51+a#mMNlkU-o1s1IT11IU1|=q7_*
z5CA~8eTV>awd2pd>$A~M{jrW9{kp-1mXBWn^`8g&!I=dK4CW}%cGuB|?Gpw5p$y_R
z@n}z>9-K#j6zhQhzL0|r4Hldmy@B|0YI?B`=Gywa9mE7}Yy4pePN_n=qrg77f>cTU
zloAjU{vEK+C-;ZnB{DD&PykrM0C@Ia1N2K-eRK}|ferK>ID_!+$$^&zuzG@**Gp$X
z{uVpB26PPqlowYkuY3K4{c;ik2K-a450BiBXbBXA^j*e@8LIiKQACFa@d%K9SI`0m
zc=!GB{W%RyLq`R?HSoyywEGIR#F(t0ta$jtbl+#h$OsPS&ju$B-v1~L4h(Q#ghyET
z1+eRvI{z<4+B)!>r)|>C0(f^O*LG&~p*Cvk2hxAPiv5o*T3O)FSAhbt{}MYY^%nGo
ze*${?rFq_^_+<<Et$g@JeD`B3-n-bl<C?nX`tb|?FUH=fR~RNx)?D^_mu=g&ZQHhO
z+qP}nwr$(?`DT)xoMiUJBz@8S13Iaso~jy%r3atu18FwbQ7{KCCz@yV|6^ML_#A7#
zl&4CsI`R`sQP6KjC_Z2{XL#}l@%RAw+owkZO$T3=gboc{|4o?2`@W#7;^^O(rwN9>
zUTXu*85w)iV`%Qfe?7kogVs#z6OwQC`R%CV#ty9iJYn)BFaSW00P#Y6C^#9WBL=w#
zD}c2CeR;1j28c(C<u6bIILGh-T!kQBsIxi&2EZ|}S4-0)U*Mw@(oPU>9`c0-=(*`z
z?Yn^q@#H5EUP^)(KZ5U?j32!_S%kUez7leenwptAitZ!~mcLf2>$X|BUqTtnHp$gU
z*~SoG_<Urwc3un%xZ3}!53|rR-4|?B{0Ak}Yo&4;PS$pJXq5$<ya6R($3&R;=$-OS
zubXx8%7lcS1RepF#3v-O3Cb$P_E?vx!y1Ek1B4A2vNx`th#}9gxG~eEt{_TXJyq^Y
zq|~SmRV-0imHW7qKvhkye%&E$#Jm%7Rw-ev`S<N@XX>UF**{^m63cjyNH57rNr#ep
z9T1GmdHF7#qKO9Micw(KXqR+&yMV$T+CZnzr7G)>j-g@1b@7f3C=%@oyjTx04N`n+
zlag>5I=Ud>`>N}Znx~uVUbjvhYRf|SlE^%uzORd*4he3&)l|;c>rwRa-eLIid8nMY
zW#3P`9?n9g*mib4+waR2W<48dko=UeNCP`FHC{5)+RR=C2Dx6pp>v~1(4laIv7^x<
ziHK)&;9>6kYqnK~RoduGM^1h_zU#!!x}7em#=W~U43nMw01)3dY_`oZ-cRG@Pta%4
zSZ(W2QS6JzFi-HEiY3@#&2g06V{WZM^cT@R!Uz3k-;d2e)~?#~-+XH%v1Ws|TZG;0
z^CmX?+sje}5b%#xi!>hyJl{0~AMkOJaWKT)Xm7MhlPt4p$lKs&7l6mKq@hlw?Czyl
zGy5;{&r?6T_gg1YD8z_@!{-;v)&_sl1M>hKtca_wjP9&d(KMVz@<Sq&C0L;8qj|UM
zBzI~ZBm<3-hp#z#Pkz2k@daNo0GdR?lp~68!pxV^ceXe1P-U*Pr~8k7**41y@sVg{
zxDnf2QXpe*t);^^(D-FDTIR_jOBi-Udli*efPP&O?j0^=8UwEWZR|*twxWf$$vho<
z*G$E-+RM*0M)q(#<x1qvuY2UZt+jT64vy-{6(Z`wnZ@x<3mkiQe4;lw-{(D<IwPTx
z+Qte=OybE8`SzPw32>+_ViK%I5P7XKgn$hh(}%;{B%`j&(~H$sSQJTt1ySz9j&&Q2
z-hWANI=??-NXY#+vP4lF=#!c{4zArbWTl3oEV*&0yWngLKg`#<3=R_%Wy@v9DWZLj
zMYe<dEujBE2nY?Kq!m&_Q!31N8g6S=0)tb#W!^7pb@<3}{?|+(fsUvgIBYt@J82LX
zA5v-)M*PVdK>#3Xl@`t2S<N>lDXZBLzNA<E{)0D6JbQeLEF@nRw8Mic2DOEx9xkSI
z7zd^eW;m{YCp;ZtE)K8*4nfNewA*rB6Y1RE+5bq*P};3IsU9N=;H4kbIkl__JN^sW
zi76us{-zeulfIC=LmKOu*`v~48-J1Pq|a_nT?tbvw&Vq*TV~!SvJ9RVih=@0JM&)P
z5nFk9pob#$FbLZ__J5%m`)oV@G6XYI34}<7dT;*(=<kH_S5PtPbs*_ucJWDw*EIzD
zN;)bYOwM<yBxrj0Ih41>O^CraA0f_nA?9;b_B6`B4?OD#qw5;#TBan%>z_sMqq33U
zk9K)Ga*jZjf|1`JnMiWY{M^cC*9uibQBd<qE+HR}6Wnb0el8%e)&4TA(I5bsspy||
zrH>7Zv2Pq>KX5A>6apEh4;~)&$Imvq7Ikf)smU0oG_$L8Io>mYCa?Q!|JK)vy=Yw5
za6B^>8T#B33wwIJB!D%1J@2&J(cPbf2rs8zt1i-v>1KRY4;EL8wQcA&Ci{=B_0nPT
zuqRHHbzsSnu{1bViJk{(^A&z6q5sBVq5T5~yb=Fn@s86JvO;*DPBfe!-ytBYBrt}6
zV6Ku=rQpQ5NN-u6?9xA^^6jN`6yaQgCkh)YX~jP*%ezZVTEd4ccURd(L()0Ls-Mr3
z__~SOzB%*x;@D$#P&k8OvkKt3b2Wn$Y~#`bds8Qc#e<yUgWofi-L!Bc)<5G8SFs?)
zT68HxFh)SH-QA+o#r+(A_oe8vL^rc9revlT>GKsM@!0t=u)khHJc^C?NP<b3li93v
z4qn*MwxlRDp(;>0@^F@WH=IoJs_@gEo8H|2dQPQM=3dg$ZR3+J^iG)0%VbGiphsE4
z@l|%@d%VSWQr&9}Gg)iWFX|@?zWZuCfVjJkaLBx6x$oG56H?a}Dd(|Lr4!+R!BYeJ
z3%H&-!DH8a{Gx4ZT8AXQvn#HG!-Ps3hii-<qsoCoux^^-Z|{3z?DYS~!Am-?CVb9K
zxWMRJ)K!lCUz?Ob^isaC&p5{I&4;~n4dc)3cFj$uSVajeWHiQ+-EOQ+50vB%=OB6<
z?0i;P0ky-#fQNCEz3H^%GWed}2ViMng3L?(s1DhJVH1Ktj@F$MgG}SS-W+uA*u0EZ
zpPKYVL6xs!IlKNyyc_27C=SDl`;|!!3XdHO9;1-XU+i>rP3u~jY4sSP8P*qi=#aD5
zIvzDj-Wv!%C_);pnOJSLDKz^6oMe8IoKUd%_U>7~c<!Lk#}bh=SNW6{ac~?=5V%F!
z5%TNCw^{YAt%))URedzK26yq;+lvp<1+|=$S2#h%`Vu1_jCK7mP}j4gh3d~L4KVf5
za$%3H{xxj=#IgTtc`q&nR<|UMp(LyD!~12EE#z{g$LZWUvM(;%bXUJEW*KUB9<~6b
zZMyIbcCGIV6Z>r*e!&=WJ>nfc6>Qp6&_ba!&EO=o?RTJ-znv4|5jeCFob3s6>DN;b
zhvPlm%ZC{z3q&ySiR?<R*22H`QmgRb!|;!Wm+pIn?o|AFKwn*E?S12sZN=a0#zCRz
zd=}aIw<*IUvKX{424We}2aqazo8XubA3C#UY7S<Jeobbx#RJDJ%J;HPS+KcRq24wq
zG<#~&OX+BE(LLlHF*hAJ&U@;1LtE;#^ePw!sU0ce;+SlbI%&MAV;MKtM3u1w#~AYt
z>&Tot_*RlN*1osqxGis!_iBhTE|SIgiU!8KC9aW;#g$8-1cx9yB553Y(WY&bNegCb
zith-MwRq>5X|cY&2Q1!NlLoT!WW7VzZPkY_*RuV0x#zpQY}wt7c!6MvakgET#Xc(L
zF;Xpq_}%}816CO_cD3AvU({&L5Pb*0SYAQ)zL^Pr1o}~^P|LIEjC2stk}^J*z1Q3c
zKMpiq63nu%Jd6fsm+U~18lfCPuMXNU@&BdsB4FwUZ&)UIO)I!gbF1Ja-O?mxWzOC5
zoJY2wITIuZVGDJ1s@`3UJX&(2B?&hMIY1ly&FI2W@)%D#*}o?frlQXE#eK$7byV9|
znkbt5JnjgE4XkqQJ4tWbm&G}o6w^pi3`<j|m?oH2p!Da}dw$`Zy_Qk1eOrMNb$q$d
z<jlJCAS0M4MA`)|B4>q++mDo7&AFB(?^&_azGEoYg7vV%_699?j1lK(dG;8HN-9}@
zBJu9^!;zR7@LlW_hOgIn%)U+_ldjpL+_PM|?`ws=u_w9^ogTYRSSy#x>jy9QAEWj}
zDnMy-;%P<hgSN$!arwGDGsY&;pA%8x7Oh>COXa7f`|!-QaiJ754<@dH0i3Y->J(hd
zACQh_Su1{x`oFVubde3}>6S^&<DUp9miNBqQ1miEvvi-efi}^1jJMZiG!K=uz`4Go
zzc=`pCK+L&@2;d%ETPka&-CtnV~aG#Pl&uXP63ITJ$U3IV5zo*%I`_G-|x1M=sFd+
zAi6rRF9y#_s6jHfP4RHEOP1d$zVhZX^}Tc(M~PyspFp}m<%aL$f}_ukK#tsFtkHtz
zL{fLGAE783c?MC2kP?|9ic=rv(05ijb<09rncKL<V(d9v#wQz-DbmIzx~|KOm(DVw
zRc+I&S|cS{S8UMJTEJN7R@|jOs<D_ZAk?nY!`VlXy}L@*Y<?gka_dekKX*>8YufxS
zwpMwSuu#_E`-<;N0aw{qnimJ=#Uz8aoVQ8cEitKW(cxMf1wK&m0@A!sGd^J*l(|nW
z*0XkXI<VR(B-KuhXJrx~BxZ_d>G0e!6qu9sp}mreRBF6MKujX}&b$EylR1NGt-2w?
zT=06!^J!^+2cXf_WyEP<^W%P%MzVVoXj-ANJ^08cBbJ^VfeT9K#m!Fg#wy*)#Y416
zRq5^BsecEnyh@-ct+o1au+%XYpzouu>`xPc;~6l|dUflvC?VB<`W)wcnqi;%kE(^D
zsu`zD7Nvk@%3gp9)^zN%6&o;iS?}oK_IxCmwHi$~^8c8eGW6d}{;9BX$a>^kCgbO%
z67`;-uaaT5ZVb}eam$hlrd=kZoOza{k(jM3bi@VS=yt~JUOe%#?`b{f?i@n4ovAJo
za}Cqzbu*$<c1Lh8U?VFXmvKBtr&7Dnp7!VHv3m;Bg@<vC4jb%6Q$4*fE_S09FumIN
z8O(#6s|}D}&GE@8dNhXg%$drS>mqd@<7gcre}f+VGZKF=`;#dklYiSZ7p;IU9es+n
zkjd+u9J30TnW*70kjjN>^5V)baw`_Y?57At+7cA%%GUwcdK32KdC5%Kbak>Ut#0;x
z%B>@dm8ULItEo`eO5dNSlf%y4#WG;{U9{RmAe;~)8u0&<?1)n|2@(^_l)93jiL#n{
zD$4Yk{oRO_V8doIH{1=YqEfBHLoKMFaVUv{^+m;xVyYXe8tzkq8&86Fji``lU&J0q
zuqCI<ePAI|0r}ld?b*udIAvK{YJBlkG(H~Y^DGUTvr!4#J=XAHaeAJ6%AlP^zXfB+
zn?kcYhz^(_50nH;Q=dbymQTo|N{F*{y>%+m1zK;qv7jp!3~S3N?CuCqT;@vPkM+jd
zV}EJ5fW-94on)-NLsOkq9zr4IO29cKDn?L3Ft&<8@3)ZC+?;h^+yM84M5%XlC}OkE
z|9}~4D!>%&h;QrK?zst@#63vUQ$<3`WfOBU$E9sF9Fq*gsD!w*!8Gaa8k?O$pEf@J
znZdv?T~lVI<Z{BB*cI$^tRjCd!ZcLtNH0*3<#|FDFT1t<kHZh{um@ARk5sbJVgwH&
zDBBw#HJv4Anb|*2A5+E`4a{j)gk;cVEOl~y5n@NREF?MpK<0jr5jU20T9(^=+6tM@
zMrYPV#VBy|ZI<Y@s6_m_$jBG{w-rF?m!GHB{$RhzE9I#UthDCt)a?Q5ks|BD*spIS
z-y&I&i9hQt<GTI)X?kX17D1Y5VU<Icc=mN*g{;6@oOL6T57-);9G=Q5Vt>@Rud)e_
zvrR7e0)3GGU21U(Nf$lOHK2lY%8spQLxsnDuB@MYYY@gaWjN-K=gc;Xx>oJb1R+~G
zzMG7AlkjCToX0a>@TgHKvk2cQ=5a2z2)$ocB>Qe(h>R{e!{U?Bh1n$wK>Pr_tL`e*
z3=Dm7!=LD@?C!`~gbh4*2LqFd#vO#)(GgF-R9_fNay-_kn>Ulk7aBS(d+(a{dwelg
zE|1Mg+F)mLFo^0%wDp-O!~f#Y8KOwUr&l%AuDJ4B*1`4420?4@t8tp2dh%$P=?Rz^
zF;1K**_P%*T)nINSi#wR1A#1mDKfUT>x=PnE&a-P)~U0ztNDe1qtR1RhuLN93twyZ
zY>+Qj&tYn$l5fJn-g*c9YL2;=+gqBbu3Y2MoFi$gMJy2&{!{Md?U?R6Pn(iGd@Qn{
zqd5KKdJ^Eqt20E9W8C7SM8KHCi#B$E=jUvI+D$A)e8WydEfsKEK5et=v}Ag-7%>dO
zF^q9m8RtWjUV)f^YP}@2Jn2k1H?4~2`=wmg!rg(y*Y8OO<d@n-K+(&A*_^w)IJC0Q
z`^{_K|FjzamHgt9T{w9_KWl=h$K!5k!CIAhU>TRnXWFDd=-PU|qt7eWB*&dVZ!N27
z-cIYsTu5HDJAU92S`{%z!;8(ebjtCN7>uWZa7dccj&9*4C%z-c&LzSH6bGL{<8%xd
z%tV8o{Uw>ril}9?8?`6Kcf6b>G4XZrv4-rFt%E0rB`VKd9_wA|w$a$Qo%cw3rweXS
z_l(3@Z|+Z9b$iM<ce(UE`j546=ZrqJWp#?XM&D-+*@%+z(MCH>ViH8nG3R7ee6MOK
z%Hp3Ssq(g1XO3XSF+}>vDt3Bp<un;%aqxSS4Xn+1^_#TMMaE=JT?)5e<!dg9r}FvK
z54Gu-q}Qpl`1JS>%KeL~*$7lm4ryGd`~xU$sc%T2;ghARlH$~^i-IC?THob$+-$e~
z67Cnv?&A;Bq6ul3_YkGdcqQE9XGjFbhWzCKoXzX~A$cms{gv*JqO!!P#!>?$qfUwf
zS^chHy=kYES@b#=UBZ7z8!KR>2d3_({+>u71#X7#ZTvuNNORUo8egXYlUl&^FOF|s
z(ACI_2rr0q*b8U3^<Ex&0*bxHkUav{^USNwB{KT)bMR5mycZeeI)8*t_bJ(Y37Pb3
zY595UP}}r_)^~WOXDaqBW1%hdnYrs@tfmZOR&+sa>w-<qz2%^lC>T&8wkgKX;dHB*
zl<R-GfEEc@pYYl4EhJq8xSTPb(JDt`wCON+Wd(NYKms{YVy#sO;175uNv1Yz&%aeg
zMbi~)X4h)&JSalz*jR3dhTO(N?#ll}7D%2X-?8+$i&u7Cq3h=d&qXd`ramvFp6C|a
z0cI1LS>iB&3>&k^K`G;{H%^ipPI1J7P!j|fg1aK&_gE>UWV|Pz45=ljq{)F9MVd07
z!-KkqQt51-_p;~U&_2%b9o?B0NFAs34zH*!Jghe6QGR>aBFqAr%G?aZyL4z8-_s-B
z8+i!S>Na`3b_6O{34Z6Oa=Jv*FiD=yz<wCYiiHAml4S_7LIny6%r@@ru*WR8ml}|C
zMeq>mL@hzRG?3h5g2K||7i{5!l1K51E=g}_C49~V=`L=w6scp%o1}f`WtS3`<8kE7
ztHljfv~IUkzLwQ6x6h-`gH1}Z%6tg{^MQicD>cnbzu0|ex}dSE<gQql`Aw7_Cn)n$
z(_0<XDow1xy!j~tO<>C5=$`*!A;SXzDDlUb1Oh?W)>~0WhX=gxEMnVss}UnqmhP+t
zO)IwP)d5xE-V6=RoGoJ3l;LV#tOn2bCZ#En1$zp=QuTaO;StkQD#fo}m_E5J67SZ}
z6h~p8w4_i*mm6z3{XMW`%pLXV1C)XX&nImyyBYKc5<g@9hideO*OL&~@#3pP6)6j)
z)ao?x1HK`4)O}|OGSCo*I`-zuDEsH;92ja<J^8MIohbqFmIS}h9(}H9ReUj%?V~)5
z`ry4rc23fqVP&k#)eh2UPr)f^#3BWEYvSn?6j>ET#+$9W$V-J=pQy4@+{B?@Z3K}$
zg8hY(8RB<>#Qbz&wt4Nt7U{nYTYERn{)cvn_ZJR;4wE!3zU0n}%U9V&ftmgue+8)_
z4l5l8g3*+}<onN)ftaPE5B54)#1qP`K<@~i6V((-L(;;JI!-<&EhaV?2a66XbaVPf
z$qmDC6izZ1{xaPWVRsNC#~wrVW-8YIKI86*C|NI%02^{*8u(z?Qx@^?P9=rdW~<ph
zViCPgTSVxLCh15_&Y|;MyOaCOEFtd-3=f_E#=7WfGUQAtC3nSh)U@$n(92<Jb1?+v
zBE*u$><fE3WL*z-aF>mA!R@v+lPsu0*>s^_J!gBxEP&HwZMk!Cn%ALw*M~D_Z=IY}
zThu2kd*IJBC^K5>p8jn@jC*1CH*H1eysvta+Hm%18FM^yrQ~jNyqvv=C=!>PS5?I6
zic4P<^DnB|*>ja*KSGsnc%$wRJW!#y*FGz<uEE(pEq{U-M6fkmlG`d%ypV6nob)_-
zgeIH0pNPzp>d5w#K6CJz#QOwjxn%hrisWS4+vYJ!4!z4JC{(*0cUDR!Rg+-MlcXRC
z%78L6S#^FARHTDnlg#LORpR?55m<@xSZb)Xr}7NUxai8zFCy_XGSj0N?{%wfZvQap
zwQJ+I;_l}UnQEgkF<T_gSUvGYGD;l(e{C~1hX2(zW2FBdql^)sft8ixe_sE$ZN|vL
z!14dvHdBITSN6c7Yh4#;Nfb`=BTO@?4UIO?5erVy5_g==KA>kZn+8pSSTG%?v=OoV
zvQ{H5Box_5G(1@qA(<d4t`kP2@d8_E-TN<E_cr~M{n2;#<#pGm=GM|x)&L6)8;U_L
zXDmI4WYNE}w3;FeB1{M>N1W);*{K^I4(zvh<m{s#059Ot+(&Q?LQg#C{2DI~ChXC#
zF%3uzg`cq~ZpuuE5Fa7@B*(AJ5Bw@mJP1_lz90RlPawPY??ehjT)aFSjF&HfcL^YV
zdS)iNbSD574HaKH99)|~fqZ&|k&)ovy>JEub)q+lU>+_d;)4Z<pt3SD?jh7eIsWiu
zKZGzrq5M7|w1EKeG*Dw|0M>Abnby#ebATE-VK#$%p|t`su|HtxKE8B1*|7F7H!3xN
zy6gVMF_7=^P&Y>X6ih$F^gr<U`{KC+(;kxm_}ylvVdFw=-r*$tZuovtcKY{(gFrNq
zsAzy;qX~a_v2ep*dE#LEe0>;t9ZnE*RxH3^7xt~Zcnbb^LIT}zdt)9%V}p7Iu6uP?
z@-*ltVZrQbUN2z)riKnFh^8snvjT<gbU8$<gNF4+R&_((fvhm$a09Y04VcnFpM{Wq
zLVglO@zbkufLH)~W~O9(seO`zp(W8o>6jOItRb-v|CkoQF+ywfub~l$0&(<k5JPs>
z)4R|*66^>NbqFPpzytW;urEX$rJ-vKjIBO^i2uMF^dE75@-H7>n*sDuK@aQ+wD;#o
z$ll!6S=)bmbFJSS=2lUnLFoS({^(ip47rNZ<o(!Q!@f9UID)2E?^$(7aWOzm8IBDc
zK9n4eslcC2;D!Lbu_vL1(;)%BXocYL6~d&i1V4Mud9ELJ)M+9gdBT5YA}|bLOg3wI
zfZt&_hlJP@FoOcq0ib^%)`4dKR#8j|3~%{rWC<a<0I_r-&Ys@k34kL)HEzShF*t>i
zEO??D{;c%Cagj_WW`m_!K;!uVB<zKwp^dE#7>KWP)$N~|VF3Z>ghbMz_oGA_{9~Z(
z8&lFVfdhl2S!fKZN!+ES^hgXML4mGSOW(p(?j>@0rO{(UGum6f$d?EyuDusi8VIp!
z$d7Wt8Y3iUv@XIAOKrnU66jb-Hy8L|lzgRp%$;gl*j|R1R??1_l+p%$Rw51paJHHh
zEiP?J)J_|Mx}sazDm7#KkUPUCdax=hzk|rRRl`587~?j}5w1{!N<<Nac-|a8aW1T;
zp!V`$eQNUE9N_Dy*Wyzv^BZh&L^NnGDrII@1h)*cQdcF(a1OMtLe(;;(0}k6ZC4X2
z=B#i}FlIzJo0`GEl;5#L9l@&qy{Kea+Co@@&~)ys{KG`T_>3S@OGJDQ3VF#*js1%y
zueki`k@2?cCLt)KYQK3e7Cnq`^PQzM+fnV*sMJy*)Es<~#--$b-9bprc+1zQ>1`(H
z7yg(=8p{VxfLnB<ZWT(pKY+UyPYZ1Nhlc%+r>8BLfu(Q9{P6TGQsZ5@ve1hmSHP`&
zt(d(nM1l*;F0MiC`fucWMpsnt3yGl1h*3ak7kQ&~>V)Ox;i4wyvHkM?#pa<TG5PYK
z%8Zk4&$7#b%gLkxx2X5lJ#k!L?=H^a3eP3l7Ag^~V5&9-G@EzQVkb})+>5L|GcKFn
z27NlFoMhyLH(S!!+#P$|4lf|`EqToncdC<JIEVb4PN~~J?n&>(R&B29rmXvRePpc}
z&#x(Q(`AA+l#L!&WJvB^6Z`SrO1_g9Vb8c}5Rb>hCE#C+Ni1QG@Ba7tCl5!$SRtHh
zu@a%+0@89-0XS(}a~_|kK(;wQhW36gksX8XZLj#&h4DyCeLg0Kg_2ji<h2%@km#gn
znN69<kfs3mt#O`FuRQ7X<XsM?5~z3FRa8~d-F#M(o{B5VM>)4d5*c<Hr)6<Z5#KA8
zdimg_`rp)l*`3{!G%6kdjj<~EJ2RI5YG6ve7p3B~$i6tC+l4#6XFtg&dd^}(B-d7r
z%c}FZmaRox)60H#jT#o0wkHct+O@8`_BzczncPgw8DXVVi?uN#Gp#Pw+E;eiZwFr0
z5Q(;~ZVobbOtMb6WvO@V5P#`|#p`}b|0-nyg&R}|d-M^f?i3&b>abiJ)cylsxRT{I
z40qJ$Xr!>NTiNNNBd%n&byQeaUP-+ed=2YdUuEh(?FFrUn||ES5YHy!-tauu;@M6(
z6lcw}spYr!e;TiGn!7e6rK@Y#y=5DTx36ex<*jsXXWLaQ5FoL1rCpCYsOTEBQF9ch
zd)H+BOx?R~<AyZNw3pRL;D$9zwCZ@>^=|<DzIId&_8i2Nvx<LRJ@C-3i}Pz&OHg24
zRAVoRymCGMjxPqf^$_+j_B8Lcth|r1RV+7v|02<)8@l20u{cAGm*q*Ux8@gEdNw~$
z<6JPdYffTjBKagxRMwEo?i9IuHbzkd_@1+?vanr{$-hxEh~7-?LK`Cxm&;tPj>lSR
z?{_zYvwy6*l78LqdXI#USgz8!RcBL)ipcy{-WnD=<5~ALw$~6b+&{j%b<Mt#(Jw9!
z#)p}hS3f_dt$!172a+FR^uj1lvFd)}Q*7`4X8ygBWO-aWO#n(&Vf(RLzd5vN+{k*o
zS7-K`$VzuqPSVY7GUC`Kh<$d1c!IM$VDLBvQo5vvd$Vt94x3>2^+4$A{_gmOwEOa1
z$eD3p!&FXjgep$icvdePJwT74`gG6IUnX$Q=uBc=g;Ilwl(fCOX+*g-{&PHv0*adZ
z{50qK;zO#QFWa!tM5SDvujZ0O)eP&6RP)tR%~oJ){gdPS&35Q|B>yroX>;SUzmz&_
zy4Cthr(}3%!O89AdPNAAGy(Pr^G#8T)iFLVW&U0r_tFU~^r1H4-3cv4#cqqgMN;%E
zcOBFJCa?J3XzTcRUp>RKx}D!0iT8Vh7UzEa)O|R)p`d{5IzzepHXvB34AYNd&Jf8`
zmp-+`Uxtg_j=6hj)}hSxQi!U~!=31mD6?dq(NRPx_oloTJ$_8^e-lv>Z>*4S8)<WC
zG0!3Q_T$|g2sp*IsWZuo#FRDmL>ltr^|h6V-v)v@au~^pG**qcd<<w-*mB-sqvEIw
zJB}3n_d>fe8My;55WB3`B5W`>9=IETXlZbL^gFcCZ_WL8@gE8m;*t~Iq~avm=UUvj
zSovmXU8Q9Q_~T)rR8oA2;<j8z5MP}XU&avj?+#|dcK5w>VaNQ%vSP5|q=OTMtmHxL
z4sBW5x`ryjDiYlUL{{g*aFb6Xt2Z=_<Aka=e+RRboUryc;jT?ItmV+TgdyNE_;@-^
zc8_`;EuoXT&Z%w{y8kxjX~G>t75o<{S^di7e^HbF=|g-wLrW-b?*CDY4ESt}|5wEm
zBO?p@|E?yP7?@f9UpWayCu(8sY~qMd_g@u=vx%^Yk)5##6fZB7le43Vfen=Vrgevr
zz3omb4fd96?s}`u*8g*IP_U@94<i4&CEe|K?fKZ|Xz9)H<*aDg70oFa#Y%`0-|F90
zUY?kb3XciTjl;?#psgAk8XJSqFE;hZCq%GkVRUG6tVdyXWI6ydPfG(*n#KZ<ngTEp
z5h1}~@DC3xOl?dKE~4;T2)@`6@$amTFHg=+V&vcIU+vqP7((v5xVfP@yEzoOxgM4M
zkx7h9!|YpHfYG-#vH*vZlUqtmNCA+O9IpT%n!cLOinaoXuo-DlA^BIQL}c_tVW9MH
z4`Aqj;eqNO>f2eq&`Hxi|AFb>*`8j1!qGdKfpht1r&q?e2X<5PO9u<6fan|BLB*$h
zlkM!TgnZ?hGa(tx`MHQe+yCHAte*Q3AN?i#j$K`U;6>nuMnLLYnjAvWvo_T?h8eu0
z*(tuB{sVaZOaBjd{JZ)DkDuau2T1<27YC(hVfrK6oLN=f9iN5ar#RF<J3qBNg^72f
zXK`u;hu~Vz1RCbnnYoskk^aN7F*Y-~dilfp^CK+Qj`=+h8s9y{1H<?;YIH<sdt_!(
zX?FYxb95rw@+<#r(;2$vyLze<{G0eKL*CDGVR5N%d->!8{j)gTJ3$X_W=0`tHvct)
z`Mp7HVsoHxX>0>1Z~q-CPiVUML;t-|Wp4bOgZV+5^fQC%fA(|XAW}1<lJoyZ^wSp_
z{Gd1IN7D0$MM8bOe<3tBJq?;~WMBwR?@(_K(5~M8{fnlJ#Aas3<hACv^eadAxAMm$
zYigQnN=k1v^^cUhU=_pWAm!JPq)BC;xoIiXl6gwHpGv8RjdNU8D76$UOpX1ml05|1
z{+&r+)VnK<OYyB{SQOIKehew8K{&y4ioMQ+SOxXhMe9yd#4aY`b4~wp-IerFpH0E7
zR^!TQAaXgWn&aclWl^^8J)R`p5SEJbYapQ&NevD6hNS$NT)wfT?_m1X!ZN{>W+<{Q
z22}TLm7fqw$o0T>ydIW1iYf>jj;J{bv#wgiSpbYBmHhX78HGNhJCp!QYn!7Tz%!M}
z5yNn1B_qM)O_kRyAeR@_m4zg!>4t=H{Q={};@u979El(mjmn~U9-1e{w)foX23}b0
zwhKi-sq*C*FtIlgYyGA{5ib?(CQWU&I+@7Vc6<0{>cZSAvSA=6^Q)>h2{iQtayih(
za;fMfngLa7wMW|yu*Ek2DjTQ~o%Q!+_d7}laA>UNiJ-aZB$GFtasp$oD*i}~+!v<a
zQQf-0^8qMCTVSnLd`QjDJ-|PC5Yn3|s+3XHESEA>9LxZ_v#AZxt?2aFcQMxk#vDwg
z+S=lA2K)fIGwD$2z5|odrc*5JTxZ-@QMgi1YDmdCl$SRd=;$*a1m2TaE2ph3+6oY)
zUhy>5{wr2{rmGMN8=sY3LFH7C!PfzXMjrCRJ}29A{ol2yy&IMQ1fW*z1j{wN)oY4v
zVGrssE)>&+JVAbI>QlxXrmX^@(c?e=f&xl}ue$3RDqDKsTc8HA!H~DKNTZe~+3JY@
ztd2dO7EbDT`9bnj7{Bs+sG@Np1d$Gs4Z-pr`j{h#63S)?=<VxEI?M}8TPQ8gTF0Z*
z5SOmbGc@iK-H}@B+4`t*cO3Shgif_&${Dwi;VyXsgFpNulp_EMT<he>Csmr#OK$ol
z+6rNOk7yYH+s>LfE4*Bb{48&M4iD2RD7H_@AFrLmP1z(`(QF7b$Uc*3JjBLGCH%w&
z@-H+K&!!5Q`Y8A`64`LtQe%GfKhIF=&35Fi&o?2;U}?yKMlk!F*&4~;o^0p*IiY91
zMe(y}ae^P&*v}!Q9B&6?nyrIA5!3gs?R1$>XA9g>B(eQC+{d@IaIJs|``w}-{(S)U
zaE)xHNVbMJ=cKnH_;v*=Y~tqNvw*HoG0A-R<&CipB+oxvRV!>djk5d(_2*iiZdn0D
zT$iz^B(bF}#4)IM%JUM1z;%H0_1UZ1$RV|`AU}e6$8fiIlmrPw<F4Y?ih0*<)O7De
zCHJq~C}8*dC5{$kMlJU3zzu$jbjYs4MhBS`QR@fA-CyEBO)0I>AcKs#BwtrSOx1VO
zQ;+^Z(`Va4HxRL^52q}JV%KBK=hKlg^wKQ3ZTZGhg(*1KW&{z2ViOi^UBqO$r#+to
zsgqvd*!=g-SCb<K1*v3qq>mDUi5Xsk*+iBCV!F_)NpKazf335eBhl&6SXI7tou7dx
z-9jX{F@ZYtK!5oR<z)5sn{XvXz}~s%(mQFTmbTS!#R{2seM~4ywKI$qPn&Sd5(%>{
zlI!5m!LGl>Xg-z@!AL@mgYfC~U}1Bp$Mt9J2!Q?!%0q?wEg9f&L;f%57tDr(U3O>3
zEJLv-(VaZEi_F=G=)|T1X<h=ulJB>r3<Jp`6$-k}Z>if!TAX^h2bWhe$F8eE1Hr%Y
z=?_IG`N`4!n7v;mS*X|^yXBHo-Rd`yy_SheZKiiIlE52G7A5eM;0LD9OR&hbdd0`s
zKM(5A^r5>xIB-`|z^wRhOv{ry|7hjvdEU6JF#Vnyphdb&G2_9pIlUbijSVq0nybul
zPvE4p#M59SK5^zyV!R!9=}NkD*9Ch}9*YTlkb};(=kiXqUm;PUTo5EP++bnBtJf<L
zA&wkDKS%xSV@;Qr7ABxTr^2VIs7JD#YJuY2{XB^Hx+dE(U@KkdL~{P4VEh~z7yj#W
zqHKRF06$Df+GVas{&YlUK+lDsd5aT+8aK_|r3gx4i1l!vQ$1)ogPt?5MB8=bzzE1t
z$QK=z7^Lu4c5u+rd=$V1!M3kT8I(3V1dh}<uy-Ad6p=SZ41)>cMFqnjR>$)v?k1f)
zvkbfLPh~i2XyRu)TR65Bww9eK`4T@>Ga0HJORahRDBXZCOtsK-VCHYpv_^~uc0A_$
zuVJQ_ZxXM#?`=hlN_~s`yi0NkNoJm99Ei6PzFIl7X;YV-Ad4Tj^j_i@^Clm}Eivl$
zh5Z%p9A&jKnLlKiP+T~0TK1B82B-yDawCMu?SX8eM6A4L^C}+!-YE)Qya3C|KE-gh
zDjQbR0Cw;5S?(9gjh=(OcA=F7K09Xn86A7;A86$2GWR6W`hw3Md~Q#mXO#$C)?1RD
z?#MRx;C!5uCv^i0U^sHwjnE3*&3!lwjYprsK<>kE7y!kj;T<>Ja)?<f#?DRXC#mMP
z+2DoPtJZIa^OXZ?t%h?nXa_nTg}VY32E)`XC(8GXa-2|jiLKrnrEq2=*UG+A64&pW
zojkG8ul8`5i<TXI^Wa887=2~6+x&n7hG>%EW5~3eJEAph1bWS5-2SPMz8q>pLfq3`
zJ7HDJ+E;J^@Ts|vFD6ik$e(KX3W765%aXqkRW<0NgSrt6&PjH!`%qwAz`$3<ef}O9
z&s_^Cl;<7%`;J2_HK~ZoVgTxQB%IhjjBuZszyKH1)ldZ?qPHC(=NmT9dSc1H<`L=w
z)R*9u_q|O^7iQ=NcDIDo^YyPsb)g)8B~?eqcPmR<3cZn?z)8MhhWu`X&ZQKHi@*D6
zR_3_AVBRRuE7Wray=EhuGd>y_%LZJLUt!mOR}Rb;9?$Rrcms-TAUb^<GX@D#hAC=~
z(gR&k9a#;(v5%m*h~)LdAgLLA!U3^#yct0Ev_+;tai%S&GS#u@&n)+(I;kwGV{_0e
zjK^7fad0P!#j@AplsiVVgi&nj263x}pKz4d5IlN8Cor69fJ_#8MPK_xqMu9FmQ*Ts
zT!3A#SgaXA;NxyT1*36s6sr+tdK>MVBUhj5S^Z1+-Vnl*CKX0+_j?wQyFTq=eRw<e
z3mM-gIn$=(XtBX~1OP`nYmJxF58K?f-9`AdyC5Mn8#^JtkX4$)eM9D@NR(X`j+eTH
zW*bt-5T?*RKeM4IE-qVU{7~BxB>pLAuSSj7t<nPRBI9Cy!s*m4oz+3~SL84tx(HqI
z)9Y|54(Otqp`ZU)4|mo7z-`q&?e}||QKY^L%)V!(P_59Ma)_{$vr2B8UKSG5`Jkq|
znbv=!(sZ)_W*6qw4sy(^^fz_>%N7{bKtlF_B&esjXXB_=<YdrsR?b+reM~87HwZJG
zFhnQU>Yy^D(|BK(64KM@kYQ|-ZK#ZX^GXz?3p|^(f<NzCIrLjO;9}kpRFDIp(juG7
zoe)+1VDogf5FYkZ`E$8pr=4dkm%^XqIAC<L#h)oGnwI4BT;I|yAjRwX-=gO0uRJ}F
z0I&9_z0~gY7reNv5~x@R3Q)7Qe64vRBT~St@b(hc4ARfem{2A0hs7&_s5{G5+cT|r
z^g%KcI<*CaxV!~_3bH+10SxLLd6li&F}(L4Q(`yjjN9@TAJ*XuGq`Aw-{tF(sdOP~
z;ijVzIm9DwbtNM3yGzyIhQ~@Af`XUMR1o{1lx5d`!w62%E!*}+HxCW{tA$YK58Y%q
z3D>L^uBU@dZ7PU>p8l!LmszG}JxB!4f^rBM9ShrsJry?a$infFdR5JIcIXYL&x)Wk
z_0QsAu!0*e0MGAet-BK=aFpQtJk;YK<}kS=L)=Q`YfwV+vlJ0mYeb|;FFa+AD&9>O
z-))5uL@BLe3@gL+&%&7^#(ki1t!zb0+YvOa4;E_<XF}kz-BP48{Mi^E%*OJ5zJcm^
zh-LYf-!~*W(Qyx=Zrs@*Q%wTKn~#niWMRP3>qySiSNv1|4_b@y%=}^Tj0_a41}06t
zJ?NCQ{Al<B5^{ZI#9Hv~OP_DF?{vGKcEvn4irYDHJB^TI`^0}1D4h-j`BlNFQsg07
zk~;F8`k8!L{HBiNj%9{h!V;y7NOJR)eGFStCkv!0sL<9qDAs`VKAp8?By}Vfe?TWt
zEICP5QYX&`p8IYrAKbAl--&m`e?G&STc@WaV(lff{;)C<8HK2zH+Pe{+h*DH{cm>e
z{jIBa5H5}|s>Tz+hZ3>Ka`&_>QM*8;6QU#U%>QA^KU);~4DabsALJ^pi)WPf^Dz?i
zM_Ce+w(?fieN!e+16~t#Zcp0-;0_JUX5`d=g~wP_5o9G^2j-u1rsPDAR73VvMU7}*
zsU32vzNxs2^A$!ImL>1oFxX0rAP2%qk4PfUvVkET9$(qZ(Ib|3qAELPDzcxZEqI9W
z`_EfQ{2_suP2m|A)Jf(IEFoG|7UF8KH>F#X^_wQ(EPAM%i>RW%EuiV=88y-aVxiY;
zFWz84&v(|Lqh8RRhsE@CK;yqRN;?FiH??!2$W4+yS^}!KekNkdQ{pcTxs71G$ufJ?
zh;=(Uz=;R-M>S!I#TcDj+lQR-8drDw){fiu@mj!#o;u&FjZ?GrLK3-Draeg|-ajjZ
zrV<75O7<rQyx>;|8*OBhsFr%xLb+PA^EPr9d6_ud`oqTw1i(IY+5~+EqVVM0<}@bU
z^VcY0_ILy<@0rbEMhQ(%g$%p$22kqJtR|Vg5t}V1uAlT0DI3>Vu|b4}U<+cPQGmQN
zPD-d>EKypu@zOl*#8G5KmAWkz1iy338=F}k7<s8UbC$miZO+!FRdu}FiZuYCxD^5e
z-r_aAYpt6n9z2yP4Wh*#(*0~ua%+`iXe8}(e8QKpG1YQt6bqiY40<JE?|@6TDOy4e
z`g5VLob<504Q?u0o%t*I_@<c#-BbtEa^zHG>Nyx{<jjrVeExf!wFlGrW#R;z$$;fj
z881FvH|V^0HiHA0aA^vUNY7+}Gt*J{2%-gV7RKNC_oYeqt=<Ez%c;rJ3{F5;g;Uft
zUKxW|5RLt)jOMwg8&j`p#q#VTd2lclLfhO^wdNH``Hy7tV)hOV-$&Z0nVHukIr23W
z*BCNd-?*kDIu&pCe)_<Xsa&n1=gilW-F|w%feICsBC+VO<Ta#l9&mHANhi=8b|Ij9
zjCjWj>S%|XyN3%;S?f7G)`(Y3VT<J=FS2n(x_FKI>?E4SuF>^{JKR_!+{!@tYG9cB
z#~Qgb!;_{jQ2U3p*o@M1BRqUkUB~#Sl2^`89`mSs?em2~LH6F$UKOo55f**mUcFGe
z7{W!&EVAzN65hqdDij4^8~<cUR`}OFTUA@hT%;Q0x6W;-3BtJW3Ic|GiZJ$w#^tN}
z&;&5LCrhs}yP?(V_IJgkVo<Cec^0KxGP#%I^WN%(rd~@tFbD>SDhP5qiZu8re)Eih
zsPX0}wpxivm|ye(*@V^YWE}*#4W(Io+V&^{)!68JrM$i1<!F)G`*BPS4#Vgxvdoc?
zlp&Pj|F+kARQP?gS@HOp(R*!#O(^M_C^j8DYP~w7O2-7T+y|72>(A{0g<k$~x#;~Q
z{Dd%TS{hNAmg6d`38%B|;syHAth3l?=3Nqy!sz~FM>te#`mi&N^l2)4>v7Tt{RAKR
zA~)$Cq%iWiK+spPx_=0pbK1#Ec!+VPIl%W~jRWX_l0pr;rJq_n@4`&A-^98zEZZ1F
zu*&{?1|)Q?e0VMy&u8dn*uLI<Lxr--etKS{@pCmogddjCQ3M$P%U|$&PjynA1to{5
zSYQkLHJ8fw*+;wzpJTnX-cQ$S28W_<%yDj;@hgy_&P|lZTmCJQhkj=1H}409m0cyb
zoz%TZHH~$!_w%AV1s70IP`4IicRgK%-28bIeLr^<cit+Ud7>)Y5Pb&Yk|okB3r~*6
z3KZqo3aBR;EL}$lw-2+B^61QvMOVlNL{G9f@Nmjgc#W#&>=7=Cq?y@}A0!zOZf4u^
zuHyU-S|MPefUXcF1SUi1RAPyOY2t#6=#jZ^&qW#TK+U7@>q9He^zHbG%Fc`P2Tdl~
zrdJ>*_MjUcyCR0y*T%Qg=GNkIg6jXaCbVpkvY?*JTi2N2p^|mO2-rJUIef(|Vl8(>
z+33LVuqm-hlEP_@G<h0R8~Rp9x@Y!AnN0)$DL#3qB)Iykvj=s2p@@t280r<0B1d2z
z<Hju#VH+*roTLFPKSSBEnyM3(MH9Hu!ot|BkgxUr9SKGAstY`=5UP0+=fHE0(w{H#
zV}F7HnrUaVsqNEQ<KHxE@kp~P{42h*=uQY<!9c^h5*R<)RAcINiN*2O`b{<h#QALl
zg7h>%26ugCgv6{{xoM1YGJIml1juyeb7m541oLslsoavMPo6b4UfGG*7%voV9kTBq
zT}}|dw*fF7&UYlppXhzjwEHdXox@vx*zY079zQW?Ja)Awc|hWF_ZddETC`!Wn|Wy~
zaXuJdw#;~>@+_3;DJUk%PbE(%IW%$R*<g?_2}<-Kl?Hp0lNQ12ED?H{x~uJ{kDWU4
z`sA#zvZ{q(=oawEviB5+btl1v7g4%FB80{4!*e-}74tnnIW#ZAxQlk_oOYpbrI*(;
zzQ(?}fSY{tHVttR8)W$wrj;y$qU_!X8KkAphHQ#GGHUrG;_W|xe9NvQpknJ*8IMI!
z!OduCYtr-h`RC2|%)r;#oI0wkitQr{kcs)|o9z{R!pP&*dnprKGcwf?D~yP~q3f5?
zI6Vv5&s6!awo{p-CCfG<{1B^J$oC;JUdL0rv}O(LF)o*^Xp@F&VI&mF%cU2aNA-G2
zxM&Sg9H@l<NnU~1{#KE#ff6Dp8KI~;IG%w3oe7&WNqFB7Om15RVF4JQd=jo^+-Jz6
z`!omkjl*K3z{rgbtv^{`o{{PEs|-#hE@W{r#0=~f4WdfiBo=>f;_1|p-(EpubqVO%
zNc`k+eMmJ0QQX?5rYmFoE2^}zH}ETAv#Ub?Wy*L`w<$C7D}fNyqrij|{SA*~n}}vh
z`%?n3iYVwS+kg#xccPnI8*iEo@Lt2!*o4D<*ui518obhzC$3oEFLW=5AawF;i;$M1
z;NH@@<~be1hYZi!7O~#JoJXVvKx2=~hf$oRFbiQ$uU`}{g&r!oYZkP_)d2-Uj!xh=
z*LFgP?pjly)>2p?y0`}RF>Td}zDj&R0n;`K3bfx7!;-Jp*dzXk(JU$8M6gt=qJX&g
zQ?-&KKUz!NcGH$X2>EmXp2^q;4K+RvQX6L*SKv<`eJG&+$uX{fud)BRL(W$K;&;wl
zyqHk;(dp;0Jbtsy3<(hvji4=h#@4tNNZhs)sJ4MpE=N80lhQbNwAQL6uFoL$k~sQC
z-785Bexx?38>qQ!j9+8+nl^cD@tYsAh?keUO@dImc^~ru6yl4D4r5>+5@L$tYJ<;*
zL*7l*2(O|K+*N5nen+V2FJH<f^Cs%mJ3?fOx5^Is7;q~O<Z?uD^cswOB;m?nSeC3A
zthrli4*%Xh?YQ@`HM$81(Ko4cvh`x)pl4s3F?a{0p9BQEzNc%u;4mPsjh_3g=bnCT
z<X+GZOJl;6UP_@nz4FM}HJM-p2nrE2%T}7wv|fM=;fE`5k+R8%83kQB-9q9#R0R{7
z_Ry5?^OhTtvYSmH4??r+)OeC<9WNyWLLUd77g*|D2+0PswxjJDciUCqtiJjkaw+)C
z2&{lfvx_L3k<ht^s~V^ddcx?WGaxhIJ1)%e7p&?zdYjojvmKS4@6x2=o5NB)>k@@q
zw$VS9*N<}ka(QtNolW45Ptel<cCehU0^Xjz4#pWHYp1Jo<DGahf_|{a!+T++y1Q<$
z1hk&h68>+o>b~eKtsbIu$uY9qlK_oJ-^<MZ0>xPtX>}u<s89~bKoqrR*@EZy)G8?@
z%Mz*UYCU5)tKHDcV2j6(ettc4hF(~=jke%kH)aGctZkX98E-+k(4!Wc?_W|UsbIo0
zse%Dlpe`8zkg`z{TNtX-eDJ80%~cn-cm_|cp)y+APqjyk0_Z7m_M+vYyF-7q1nPnc
z$&tGVyg4V8N*F;h?@RXs8LO8P5X5ZoR~{&jwmaT#R=YB+jH;Qpf7(=^<{p#4qjw%z
zVj|QEFzDlWl}{<tG_cj%!`br^?#t-UBC-`3Bs?vtiXB}!z;^w(|Ct6V$oh?V+?r?>
zUK6b4uIWe+@|0F@|8)2XaMjOp5DqA#A^wnEVh_zDMXS+3R$#BuPw5-1U89yj5m>Wp
z+fmkF1kUxrXvu_&z5S9y1AE#u^Mp&bkf%#0f2(vk$@#smtH>*_pgUsm3cJsV_ph<y
zW1XraBmprEX5OHMsM)g#^{mCpA$Ul&Ht2Fzczo{UJ%|t_xYUMZxVdZ3$80SanB@mU
zZ5SdfC8wZ`4<e+IKHNGE^@XZ$VlA_>&jZLddcM3!v!sk!@*H?bq?fbqXb0wgMhlcj
z0X_ZZSpCAQjO8GeX+E;`Cm87Ue$VX$;X7|pq8e5i{KTXZ6v*7o*{R`*(fUazF-hnV
z4L|=ZRfN>AE94O+i~SbfJ7>hT_%|;Ka^gd%=FF5sSp0zX&6(%*zFCVXx*Mlute9if
zjd6-TXuwR7F%%{hjW9IbeVZyFtchLGsgCshfNf@e(<4oS?5h30nRtQq`C-722XH;Q
ziT&4BG!E)&<_rSqEw)8rFyK7iny0jMjfHooTTFdYh#ggYS(GEdv=Kq-gofv|+i4ch
zz+A1<IjNIf?*=~E2kzmQISbE?6dKQ|jR#VD6aAiw2*zb&23PlPsang}WKJm_yf~~1
z$SxdUR#3aB_D7#!N4st^ibQ)%;!d{ujGS|7Z17>>oSGg-dU$WwwS47*>zkg*YwT~l
zNXo-DDR8d3-{c5OG4A3$wd4e!yezi8YOit7>3fqDDmsJ<-WV7PTf?h}_5iFJ<Df%d
zVKymfQzytP8elKVM~Y-`+GMclyvIqp2<YehTeGocdxTW$2luE%cG>E%dG%;;(uaE1
zJO$M&jt%~5Mq-gF*ePQ<e`E3Z(`u_>;LuYAb$!@$)>}Q(p9grZq`nQ2I~b7Le3<w5
zmVUZu+IUiAZ}*rNe{`~N9BKfCaA?9#!rrgB5+Sik$lkt8cnsuKjGEnYXYi0585!2O
zv{|3lEeEOM`7AR&`&Y_x+npZeeF&`*!MIhaX_td&55}Ey2gv`7@3XyXggq4?pLS{V
zeGvi^GjuWRrS@>U1cRdd9Ln*X2JCTAC>0vqxHV~JU~MDnijSj6#7$gCUnuh?J|BM`
z(&d+pMejS53QMGK<a19oSo-I({_#dX7e&0m5+;6Y?#7LrvAA!RQrY+j${s0=T>EzH
z%5#ZOr9NC=<sa?RyA3I30FK!-zo|#aT%IS#ALZUu>D1nkRSlm4lA0IdCrDHgws|DH
zKjX+;GR)l6Oc?1mz>Z1iib0|ExGF!7R_1I_nffYBMEg1$Onvl?ToxM!^TR2dWQ%Bl
zOHrXyvxAT|@@H<EhFNx{5lN+s#n%!Zg8#$VIrRv`0Lgl6+qP|c-mz`lwr$(CZQHhO
z`+Paqo9zCGPIs!ho=SLu`Bp>_k0$qF#V74--bVXODiIMKw!~G4&MLHnUUgdymjhKr
zE=WxDo|TXpxpw4m_@iNvnuL^(#o)ls`1E_tDAGd#+>);bGRP>|jIx~1`%W#CA{O`$
z5FfXlyJ)_;RT4tA2uV8R@1>n%a65$V*~Y~-$H#t$9^83ESHVmr*+$D~??gp^^Q7k)
z$8(-HHH*%p-*@#h(`9UXiwcu;qTFqAzH%)myQQ{!B$++jE!yi;Q4e3mFHk>gH?zEP
z>#W4Ym7Nqdco8nNUmFXKn|uQ&T3&_qWTJp!A=|K+{!w8TTO7vnpFom|+AVw25a=+4
zL$~OUD_R&M-|Ep-t_!zluGjRiC`HNjPa#BAjSX{;f&7E~P);^RE~<+?g@IL<ksPvQ
z^dm%p02+&9=RJ-m)=plg88ko_Kd<3qj)Q9VHGN%8;6A937)PG7hr6E8UDt2DZ~x%^
za^}pY(ZEZI$UYhPq>vQXpDx(;3^j=fWh<HiUb1dftH<zSVnCg})=q{Z75qLOuSv|e
zz$A0ne!VWxB6Z`vV&_rRkS>!}L&#BPA7;IWx5&p=^yP28e)f4qMrjO~2eQ5CqY~V>
zIsSGArvWL=kmx2)PTR9qf^EN}%l-fzW)JpWQuqaseyQAz&)ZuUypUX{26%MbP#u}n
zgM;rOAZd@zu?Vg$W9y8eH~<58(FcTT8ytp}@IM_!vJk$6Ku+c5^y)F;L{DGXNo4~$
z3eQ%R_$$WLeiy4qtzP(@H#3pUAPGjH-wY|GkWFGH(b2HMb5=C{t;P`ebD`;}%8%Nz
z`L0vp(K0-CLpu;lxPIXtQO0>=A$ioil+z%%$D17U(mbAz#L!+JWHp|v7%2Bb&X#<;
zA)|QPcOW6zm<=9pAx8t<BL8&HaL9`0`oeaV^OI*;`N+riJiB51?-b&1^(gMtPc=K5
zA7RdhMMlY)BXQWh>;dn(o?XdEgz|!pGFD*D2a{Km^;kl9pitefgr|(74<bzdAnlP&
z)JTb$>660l(?Lc6j)FVwtd+9pGirp&dcu#|bPD#))AK3+Z0)k@>_#nX2L*ne&FW_@
zZayE4v3xo{SqChvL%e+SU=zj0z=8tWuhWCzapGNzIqttuPVk<)@@JiCF9!9L2x*q_
z7Wsr${o7qorQ+<6l;0$qs$uJA_xtXtC}tx)e(NVawcisS$y`gsn@}D7R{z@NW~%L!
zALjvYy)#_lVclFD25)3xyb%=bft?VGbEIRThT6)1eA;D7d@IAR)8WK@>do5e(>ThY
z9np_*+>y4fzmxwGa>6?vv5g^vtS)+{ma@0audv;o2H#Uq=Wbn6ndFNgy(Nw~y|^>V
zkHn9MX5ZfeD-#3LNZCfU6xdDAUfuBJ9h9eIhF>g%8#ALLT-l+Z)A#|hGlc@@aUMWs
z;}W0FAlt`3^_25CUbL;*69$#aZ+PG=4E9bCny|+(ab68Uaf@yvLf(k7^RUG=%U{Of
zwbi_P<=)7g{>j9Of-KepBY25vIw4ZZ5Kec<2$bHwAg5O?2Ug0^$I>AtEVCD@>=U4x
z&zHM=)RSoo;(%ps&%<KA8X3je)8dx`qvYNxeWea2M>-3Hkl8FS@~cQlD$)?rMid9v
zCE2VT>}>3GIMd=&UNWi~*zbl<!lTLwa%Qq+6yRicduY#$y!0h@gVrtkynNYM&5V&!
zPV`=}Y~@aXXhlvF$}FxoRr#0xy>HWddfT2vx+&ltk)1}njWgT&G!1d)j$&5YgOR6d
zN<mg-dm3s_GIlo)XVr3rgd<QA&ps6_XUYko!%!kXW5e|B3&hEyT0~F?KFV?tYX-iB
zF+i?Oif%zcl85%JeQb>Lr~8xs2wM=5FAiW_)mvDd8QiQAShjoj<*vs(2dz6x0@RE%
z>5-8=;Kf#VrLy+qu<Tm745bQN_EwV0WqxGBOB0R_Q6nXpK6HdjQ1yW5^{d<gb&C7?
zdw@NS>x3HGeC4jsFw@^TWePmX&F5#02!juRA4(aMceg^G#fdJlPJ-wrpA1?b>UuDf
zaVN0C$V68#J6%1UmlyG3E5-@zWf3^&1yP;>$Rd`W^=eEoS_QLK?Y$a9nQ2gsK5%&G
zBwz$+kW#Sv-JdJ;$!dxE+5ta^YY=G}wPXi307!h7tBex#X%8Sd#L9OcuC%_W>)mQe
zAnU=TfwT+RKXVu%w)k9sFs^SBVleg}O$zR0&1gcFs}h7mp#^%AmE4^vx5`=8^t^(R
zDF8s^I&?3cdtR={nUs`-=(IHiV7XI#G_*4hgHSei&9xgXW%_E*WQ5QLlZ$YKl-a8F
zOs(OF3Xc{UJp$=K_P7N?66Ny&OJtErXA%5{KbHfcUld0x#iJ2rNn}`7y-w*pT02I=
zs-!So`xE4p&zOB<T5<9$=im8>neF+7)>Vxj_}UAatH}=a2|Wc;K<<$UtpY#Jps!s{
z=+H<HSkc%oGE1tvr=RaqwXwetK+WT2_(8nEkIs+5c2`*ddmNED8ZDv2cM9a8k8kF(
zFaqUrqNvC)H4C3ct-pvLHyZn6SIh1rgpp{2o^o3vM-nH%d`0Vt@%Gm3Mt)zl$H~Z-
z9mCbJnb{V&<x(F#b4ZUv7TdMvPYzn}Y|YbaH7YXwip1bf36qB>lvDnv9DN)Ni^Hg8
z1)QNs0&xJ-W3IJtfSW(G-}DvkVsQT)aowJ_b=K`bsqQ_3!iNs0{ZSlLQrttV>f*wn
zHdJ!-lQK)Tg)fLGqlgMmsfW;zDNrMl`5`lxfhH*6g7l-{N<fuy)nT(bRf4a+<Q2=6
z;Tj|0AbBUFd#Csd&fJu=!##UMi+QP2;$wUdPXJrDsG>@sN&7^-7oQ1jKOj8szP77k
zHlCBbvc&3>M>F^FQ5N<ZrdBeex|BqLkQe~qV@J8RZb2JJ+eMd=_daj6M%b!ER=W)?
z?E?FEuP?fO={9JlAEzzEWHBh_izqG=HCr~N)HZ_{;(Jg&jQI1}I&jJbMTD_oFu#xE
z69JnyH~@QQY<}|RXyEf48V35a8_KHVGi^O{@}>60GOc;oOdAq=`V5x%mX*WYMv0*X
z#8uS6^39n>A-TXQ=(I(TM~Si2+OJx^0N9^xs+XOX5NuZ>dZ94Yxt$JF2($m&9lT2#
zjrH~zxVllV3&kNa5MX-`OH*X9z<pLt%in4?Y<>Sq%T>*1USC~tR&l|?e!~O2lHs8`
zJdpqUs32ox5f|#IxOs?Y?!hyl#4@OHoOqZlqrm^EjHOiMV^nVXl<54-;aL**v0ln!
zWJ?Qn>3-CGE~aUHgb_@95kk}lqld*#CIU}ka!1bJOPuF+d$f&Yyd%)@kp}@y;Rq`D
zWpqDOw6`+fX3HP|j*t%gY(~C{wC_VrnCHgLL>>NUjBo4y0~@YIHo;B=9B=bnUdR>J
z&w<6n(u-zcSBp3jX)<O6H!ZvL^udu%2#!3Ev57w<hFI^Eu4NUaUlcw;uOK8fk|0E9
zN_$Uo_TK+m5DqDKP3!8zop77@D;(!szzDvb9s}GHZG7gTR)gb!NPCK3A&lEHYY6l_
ze-(E_w2Ms*-Tl(x*?6r$+%!7`XyGBcuPs*42f*hnnRIFI1_;DItm8LWIhG)-IIVOE
z0Qe`x9Z;CdJ4bsJ{XjfRCjGniU08{!pWz@Lb7L%it5Lj76(VxQd|#3H+vD~Cz5`MT
z=^LvX?Xv!DdyNn?LsCV0=A3V&`iz^ZK=khBYCz{j8Xafc{Q}%%xo()E9O=#d5Z64B
z$9^A|rXi*Gphu=IY3s2w8D4pQ1#SMKP9X;c<;`54pbm!2E1;fFW08$*Z*a>LyY<?T
z<URU@a>fJOZa_2U9;ViJ1lK{HWv1x&GtrL{FB~1nUMp-EWc42@)7c|=J-f(B#U};=
zV@l~)chwx;+S97`q&h6(%q8x1<0Te5w%5yBMXh4S->5v;7k9?!-D)ir-CXKRlNo_|
zi!XtI0^fOA!PYWp(igp6+raOhH8Ga6#l93TO8@V$LMCK$DP~Bn*@kZ`U&p^wMxvQB
zt%i?DFH@8JikxMAz>8vG-4bc&2q0E?^(z?41`IR6xo=bQ{kCl4wp#ip<cBMKEDaoH
z@fU6`unOE@`*CJXC~<YM^>K1S9`hO=7CzpkH2+RZ#y|R&?IF9^mj4DazZCUOL?Kf^
z2$TJ2aG5z%!q4+)P3Db)AwQf<`dfrWi@*I63B9}~_8n_T-4@^%7|o2wC(oK52;0WJ
z3d6Blab=Psf*O@1k@z<GWH`nVI$c$%SM#4(;pPffEW_*CDUZ(+{ez-{3kEmDvMwQK
z>+Lr-V&#8UkLgoBfcTS^M#?F@N&evgc-)#JRoL8cIY*Z7YlNhDw(~K0F8AHD#R;3A
zQ|{A7xL3>{om*rJ+38t3^4r?9r`>>zZtRd0hf10@GTYSN%l=5EaW5t%wQ1E_E6gUu
zL76kzTAsbG{&Vi2*M*jhx?^Ed4xZ3-oB$>UP-6*aWiMaqa=4!GjGo2V5QPQFo{;a4
z(BnfA*ZEm!_55w}FAk99r5K}=7sKRSBA35mTN&+}IfRC%zre_0doj!=+3{_Bt45i&
zx#}l*EI^Anm?dedoW8ucd9p*Sx(i~igVt+6HTU}{nV0PT9>RbJH<14U?h4X8O|c?-
z^Do3RH{LMqZpbMlF^@`XY56P(G1U7yz{z_|2Cy-r>NAP<yPA^7gr=tL-Xy-YuN%cI
zIPj3n#a8dEv8l;DQ~#I0y9jFQ{xsH4U1L{|8MkP$Z7yviSq)Cjq^4CV((2SY7-tJY
zC4#3<7~xJ^E_}JAcyQpN-KZH?-MZNFMOxNR(Gyb>h`41HMDGYH_I_^WdWTxl;LC8j
z(Mw|^O%qi2&AXTPXW<d}#i7yII^iF*I1V?0*2`jQbEM9LcTL|nS3LLrD0j*~8}0^|
zKv;*jFsAT^VM|VybQ{O(nx3DDRP|<-9NE<u;eZ|-Bb_fr3%T!@V>6<2uOz(|I6jbm
zT_UN})Jwkd^Cr8!!XgO!K}(cl5!^B{1!wW^QVDclnIrmtSUrY*SM0b7%oo>L#SqEp
zeSjVVh1DTS&@tMQqSrZjj4d{fEr;(i@-bf{xp?nW7{Ts!Dn%9Dm_R1SOrv-*x~OZa
z0$8*Em1zu}iT>)2++=0n+@3@^Ur4eH+cT}*L0V<~;ph&)3)T0B>F+zEuT2sW;c>Y5
zdP7bb#?dU9Fr(zN^~Bf@Xjodl>Cx9Yomo{okAB@?uznn1P8p_fE2*)z7T3v^U^xmQ
z#Gq0=h|QtEV<8gjDTx#78Es}E5{KU)!M?~+a-O_polsH1s@O!BH3i%RPJxa@vcl#b
z2A{%o(9L!O!To)MoD9Hv=HgKDx)>#wxs=vn$ZBaukgU!rBoCyt3SCjsiXb#Vzx{W+
zuHX3YiSf0KvSpKGe5|XC%@RB9hX-WI(&bcZrcICdCA`;Ck1$E((noBtqn4WBI;`+#
zTfGPn$pSw5CuQ&QYj-MGU;VDf6Vx=Nz!=AtbOxNz3&b*{K5Jg0cZpz<`}^SOh>lZ2
z9=7ozUaW%1vm?-*k*E70isvf7s5Bi;|HYu1x*kt)Hhp#>unM1Z9*=(`M&eNd4=KZl
z!L#+QHJqMCA_KAua&h(`Tf74mwyWJ|=$k<o>ow`I8Q2bOS#cYX!XKw0uoa}1&aR+Y
zar9Q}g*aq{5=k_SsF=JD&3m0kO^i=$r^ggd$Hjeap#y>u2&#3%ip40*MZy6c^B97#
zJ<h@95+pQc7h9+;iGAF^(Xy#>dB-#?4$H*CM>TN_<A0Hno-w@82`4e_{<mI`M{o;k
zRwjA#qWcxJa%;E{M<S1M?MyCWaoLrBWe_+kO<yj%cYoNOp1m8j#lp%?!LgTV2=_vP
z{oHcRn)g-tb8-Khtv;CQf1t_Wnt`%xgt#xu;YF`cgggFG=@>eYfMza#KAfvYa;;N;
zy=ioAg<m-WlU*2j|JfAPuQ3<K+{>&Ae^uI0_+D-V?$D2Sh8_ME?Oy_eKB^LWrXU`W
zz4TlOtgp@rVujQ2^7AOCx<>z5*>K(z8@DD)znUrr-Z$SE|0I|?Q4a{R*ph%{$JAI|
zwg^<sAz&Gz1J^z;d4#zPZL7&vwLOJyX+RGVp>?h_KIfjXfJh&=S|5@a6`0hYd>Ju|
z#)qQB)bkb3n_46C<>ju62k%JPxAIZ~IQL*Ce>uQtx`-5}$B?~@2P43Gzl5KGpTg0z
zT0dOsIHEeZg$`m8ii`Iv0`MDIMW|8l4p8#Zjrz0|!8>q(@!bB(RfWc=UwSA=-&Ng0
zIofh>DCd>?!gC#0_u;>{>Q*MQNPqEf2t<7glh+(7sB~44L%MRTbc*hzQb{9-=$G;m
z9PNF_<I`W?-_+OFx5IkTcNJjb+hbd1xLoO<eUcD&Gz&_5E&Vf=N0TzRzVk$cL5Cpz
z36_s&Z@pgT{8+F55LiRLMOvXNewi?}Ra}r5oJDe%zT+vddBR;%2HrRSFx~ai5YkSh
zGs2oCAU^ei#tiJ5Gkj*>O!iwvJ;rYxeoP8gtbIJ&k!QPWFx4jh`iJD<WxvN&)nHSY
z*a<aDBX~W0!V@Q+cd&W?yLw>89@SPdrDt_L(NBNBqjcK{{GJ0jl*W~79iFs>kS%8)
z`eC+A-&6fqZnrXJGA{utt`FEn(6**dM^`Eg%-jM3*;RVrRj#QVA$T=%(=e{{u&f6z
zcklVeR*h(o`8n`xx_IDFhxB~c;hW)}^&~BAib?`eg`iQE6+6|axM>6$8fZ6G>afrZ
zKenOtbi@=q5g<H9M5fWDO!B#7%$y?moeXj_!0;-NfXPK#Zp7u?k@_#%?<4PZMH+8V
z*Yl7ah~$lbV)uQnXg-&>{ZSY2XlZ*DHeJe&nA2Q&t$1W<QqPUmtbP(s{`ZAf2X|fE
zq{Et+D@(v=bCKl>@bIymjKHYz1h<z;j#LAE;WIj$5ceY778zGHB5nKzFN~LDjt2<~
zv1{+O7X!33Ai*Ntlcon~Sv0N(8djr6IVX`Y=wo~x$8U#GAKTfpq7c5$^hN_B>pUSM
znr&KkQ9FF@Qp&|RHw|28hg#g}LrT#3MFf%@^n6^)uVtQVPjXG8euQ+yK{-S%614HL
zZ?&>^86_;!+nMbLR`hq+MjwyU>+i?Df)=!EkJ@mf-@ghpv(S)ZaSVSqcwrEm;1uq(
z<J?I|c-<ZEXCPl>pr@BV(~;V3{PGpe!=wK1u!kUvgF<$(qQ0cvgfVxi@WOG$v#mvh
zAbI93$z|N?RD^)^LaUdeOAM_I>_<9*6qBYwW@(>gKstg^oq<&OE<$tp@C!CQxof7H
zTs$PJcg>wr4&Rt|ut|gQQAyTObctRHz7pBDW046Xae2@?VX<>EBJdr_VwKEYNW%G6
zTtxMs!+LZe{8KiaWVyKL-|{KJfz#yvWJaK1mN2kT#ey^6te~)OP)G-U2y`N~*8T2)
zyfN^*jV7g^(v^aoRe(0mDfx&J1DsVP>Y%Y5YKT3q@IP4nn@pB68&+z`<qM&)7qRB@
z-}SO>?$@VO8VMGnXs1Fi&N$SBGjFHcuk4Y}*fNM{?}nzaklpb>q`B`?U-i;HY}07@
zLv6n5T10P67mjz7ITwGEwN4c~IDp3hznn75WAg9Ov~q>s7^~6W`9*<r7u&&BCoEvU
zJE<Han85RN)TZv*KVq&%J0OuU)b~Z9Olw-_q4&_MkgI+1`|6FNlNI+%iat^!+zDD~
z^%ku0wYPYpDB31r?zM8@ojc4%q~}blTV#=h*6fAnZdtDXF1VNZ8drI#Xw5}Zk;IQQ
zSac7QZ8l3>-hKxnbN*%98({4oICIks@(<-c&4dM;nWcghx}xZ`XBAPQ`xxU1B1m;0
zlcl-nlMp`O+YO!E@)2r=!`^}!4w&XNYtr!7#W$j)JxSpzr{-Ct5(`9)22N+n3^qM(
z`yvtC!(&ck|8aHb&Jb!;5SAjS$xn8XWdZy_h=!di3kfqMmCPUL@<IU#$IZgb*#^;O
zt)M7z`aWbq1!U2RXd=@Bw9}ffS|j9Lura+%)VdZU&_O3bIzX7PbtpGED<4FSyBqc4
zCac3r3~$p7ctP=+vSu4(i+dMR_2up7#g!2Gabt*{rQEv&0BE^<dR<%YfDZlx!P62b
z-xFQj9L%B=);`N_(83#j)ke}hPP(+8ET0Of?wEY>=Fo7f9$>$W)1&^~a+S`K+RMqp
zOJH_(X5-WTCf~=E)}AV^Z=}wZeWIVn5_xrQ+TxBb(+0P+t}2BD@NOxZ%0`7!h2P}q
zl_-jGI7y>IjnCsI1Fc;-@KtVFJn{%Lp469hS}G;A3I*)LWLE4*{(Ovt1R$TTgRu4@
z^<QPk?o_AiI-{c(@FMShltskTB6z8|eNG(S=Dvy!S+2pSY?Sn{A%+UHjcKt*S_5q{
z*8%cxstlJ518nS^vHZr*&dRbM1`xH^*na!358i)ZVQ>H8jn>s=Ak_FnhDZG_W$lVk
zc=rAtGWyDsUqmq+4H;g9dIb)g$Gl(lnWO6X+gd@jFcAE&TD`yzyNaxPyQGT=(NENL
zT+mHOUXs1Mbf>w_7=*h9cH2V>;4b;>0|l>ngtSqe)fTHn7uAQ~Y$z$9ISKBwDCpCH
zkB0m+XX$+R<!D>cx(KrIGBSncB<<KjxkbckYM24mi%7EROo-1YY;_B6#Z>AXdoe%S
zIrjtut&CVnOY=)+o-Jv+U^EjFueH~pT;vPm^cSlMYMykYlkMZ=e;ADSIYIyrDhV*R
z7~4NV&Lvn0&;0P|0CO83%<n5a&Fu-a6MHCN#27yrFL;_83}MAbz(i^F_I|sSRQ5NO
zAwWrx(3D1BXr_7e5ymNcT$^2fr_qr}V7k9%iZkIO=mYTak0)#esK>6KCl{6y2?;&L
zoY<%SjZ+gHr3ccT?={EQyRGvGx{@HyUyjB;lHR{Kaln#w3$_BC!C+3wx>+=R?W{r#
zo?xk-5HO_;5A$Rz`GvEJA51O`c2d~VX<oFL7jI4cqgm&D77EK3I-YvH4Zew_W(DH#
zN0Zu!j2$^?%;X{_J&6D{nmqN&bQm0I!#?h5_Foe}C}aqy@MV31JQc78p~{a;$0$sn
zu@2i?xA52gw~e;^kCxlf#txgvUJ2bP&mRNi)L)G0Au7&lJro<mO0hY^pYMkVKjs<V
zmQ(aK8ZfQ4w-;%oYzenCOtMZ#HQuwiz>$T_8LF7s@2$>-Y1Oyv-zZdNCM11iQlqn^
zP)r5uZhtT+KMKign}f(i2=;tVe5slX@UKu@NXpx<7zD&2e;BhM7VZ#7{{9X1GnXPV
z`gS{ZLaddsNYl#bU=Z?P?vqMbR!vr4vg(V;fj-ckNt5%uJi+<+{iqz10(Cl;)@rAJ
zsW&Q<U02c5hUp1xdio3-gLj8kc|$b-O##C{oq@w>ga=t~#q%!OT$QF~3PXzU92x|d
z>K<c_+lYK;Q|t$gHyHzR6;{9?-UT2kkp8-|+h%7d(*NAxD*X}C@GORgW&@8rnLXX0
zZa9Y^YGr@TATZ<)YW81EuHp=)$H38OoSHw;#`)!X#0Ju(y${rq^#pUQ)Myz$j~id_
zq0JWwcsR7@#F8rntt!8Z=DG98gTI!myt<}ce-M6^%4p2cIHsr`-Iw`3Xr<91$P|@l
z?vY~oa{B<Jm6~-@nQ>^yp90_0h4b|C!*m&g-T}BbtSJU<XpR_%fOBah65<uZ11&}c
zf6!v)^Ao9Ayfq{ELHB)zUKR(zLmP?fu=p{1!)914tO?5DM($;<g{w6|XM9N6l(M7=
z`+zNdjSw_6oZ5QjphG=_*<!(f&Xz|;Dri_mna2#(OJWK<Hr|kUjMZ|iqAIDA-)ush
zsx=R?$tFd6F$V9t0j;h!5gZ0rmU#xv)1M}9zI;?BZOkVYt@~8(-LN0_MeMzFt5SS9
zwdtGE`c5q0Zt&z}s&Fh?TrUHIp@XK}UrQ^cuipkJxaM>c7tP<p<OQPmUz^pqF|P9y
zZrR~$JCxjM{Z_<PSLn}pRuj9?AKxRP{(;iGr||-F!$E4s-@KlZ`qqC-j+)yVR4?fz
z6<sPcQ{^(%7*MWZsfiSgmGaDXAhiYu5f-DuJ*n|oGZNF`djBb3Ev?kImM4VsPH&j$
z?8GLP$PAGOct{z~q0mx(r5O=az}J!<dS(Qc=Zhx_WTS*Hsa|l$F#l9)k@7+ijfzO-
zp%oD>^&L@S5PWKmO!KSLZ){U<`r(;Ictm)WSb~m^5>4pDE??i>2#9QmB*GB%PTbH&
zmDu;>{g9IG)Sw-j-Gn{V6rrZYZcqoHeA+9y!dQ|zc}gm%jNC~g6P!<E-|9N&W)QSg
zU=4-pjS#&F=%83llPs6$Fv9$Ws(bTE0*bD!Wy`RS9X9vwT5%cGrn#XyG;b&w(0hlS
zE)!6nxs-H*Q0%(W9cdSwbHIr>LCc%ikI@Lg)s+yI5abuvag4~Acvd6cI|k%_@=?r-
zj7HIimDc`hca8_lO@U(&$#7j{T3e$S{mlSB9-i?tr~!kljkL$K`eHqwFO`)>Q&cDV
z@At(pJJ(Gc%ejk6Up1vG+87@IwYTOJc+KS`Ac6)r6noG9gN7}jbVpQ-jJ)sq_86~;
zo#A*WB!V$V^~1rlwnt7{guCU+TBHG&Kf)Zj*|-Uf_X2SVGwnF}w5zFf_(=c8znSQc
zXHWH~A2y9{j4J)}3)ZJUBWOJeTQ$L&Kv&NkS!rD^U!DT3fwXgK)MxwvofnPt`4;kB
z3u?^D8u9ZNMps;hIGOvY?lBjrn$_vGp24UYZ)J;3bDnO8Y<p}??Gca2Z0S@lg#;#x
zmh$JWdU#`oYguZ)Q)NeW5?8fsizv#~-ORrmy*AHn6yqXlM4$=o?5l(e)p)B9^LD{f
zKAarx1-zJ3V&(wddHTu3uU>?{O}44f^(Uqz(Am7*402Hu{Wjk&mc6<x@A--LC3j2o
zjI-6JU{&V58=&nx)Vz?@tuNorbX1>EXa1@0QwcPD_dH6QB%Q>FiIm(w%y)Ybr|~-y
z?pKu|1~<~_?NewgZU_SRy1PDsXab|c{fex+l~08IWAK`s0O{6J#p3t*S+1CE08EjE
zx)&pcte~YQa%e9~gO-hWZt`}8OW%FOalW?3sm1lxb{)Xb_zE2+=~Q>udefIV*6%q<
z10wHa8eztXhDt^h2vU>Oe~C?mU!Sn5b-Q+r^|+grHgF3QZgAp2j+{x(%TW6hP#ap?
zTu6bb?8)X1#LQwN@ABg;SrCWO>#N|@6+uB-$tVKeV}cKz?u0Q~)Y<v+fe+Cw42Ik!
zQy+8J)FP61Eb~JN6cX0bC*5|k)*(p=1e%6`-N>)!V^#$pb4NFYrs5Bd5Ze%5>{Wj_
zV%zMhQF-7A8PUw0L{IJD_Qe{31rvG5&OrMK$2v8(FAy&JInZDAWHOnrjGkY#7a@VP
z+kA3)dyDAGoqJJ9T3M=6E}<SY(EUk@6(+O|t}Q2p7f8@5qk(_+R#6stVBf4QfVe1-
zidzjvdj0!&?kG7ktn?;S>39oCErPdM5$~eWgl^K%nG(IO5ilDJ9&1}oYlz4$mu(Ln
z3vOKKH7$1lS`4?4;4~(l1bu<@vk|$$oa{=Tkk-ye7xVkt+`=mNgMwqQ95ZW!{`$JH
z{rUs##1uW1Z_J5(^dxgFjc#pmrd($LjBptJFkw*R;fk9I=9)*LLIk1-bzV6#)>mK-
z+B&X3SfHIkvSX@UvbvH-eA7T!SjaVn&<&?WNfL1UP)nK9XDU~fya0?*dsoRnnD#r*
zfpK`V6Ebi{2~pis%aLdNIk#0oKbG=%KRy${t#j{;CwV`*K_^(*{)bUeFw}{=KDygf
zZE!zO=}rBc!C91u+>gv2Zg=w&X0@+<rjpa#b=Nj9XiJEAuhGXbg7J#y#6r#WM9MYo
zEDg)(i{@oy@B65E+~s!Oh^jse48?RA`%=%5T4lRQ33_LG*28YP(LV2`q=(}^7zHNX
zcAw@lve%;hUr`h-5O@OMNBuNyUkj_`5BaFo0sl1yyJL8r@75@92At@xp~{osZdf|6
zo`sno+}f}CsO~lQ`)zlEyy;LbgqcG&Fjr@)0C9%ug_kQ6sK<$%^hLmH&V($ctJ4{h
zWS;pSGk$$Crt&LFm}?bjV1%{z!NFC!HLiaFh{rJXEW{>~4Y9t#;>Om&)en=>z+8lE
zj_PsZQ$ivY@mB&s8!=<PkoI>3`tnFyRX6y}!5HRF=&#uE{_tOy@!9mOlzIZb%Un9i
zNUaP8DV^X-zq}gA5|L#>WZ42aXjwW8(uPCGOZ11_R7Mk#p*LqL&;3t%gysi$;n<c@
ze+ic;WT}CQHoNIAJqPy7*GyvQ)3EMcmp?IHQe=ZO>iRvg$a5hNW7FoRl5c$?1~J_j
zNLPy>!@Y?)I<e>gWh!4ee~Z^btH;6Bd`uT~Ofv*d&x6_Y>)jjoae=Gbu`U8oUA-ty
zq92@L8$+fUvX8A6R$HZ(4g|XR-#eOJ2I+_DV%uOLnrz0{Y`gQ;bM3s9nHsGSt<*w>
z4|}HZ?*OmZH)|tX2->a=V`=b{3D-E?wrFyYYzR<kK~{?#m&j->S8%(~G(fAomxCOu
zY&Hp}R++usP&Mzdm@=KqVQ0y&l4W%l&~8+1jEH6W>q1i4dys7IODjjE0#vR%k<!uZ
zIR6A9%}^dNoA_Fn@Ir$j9s`JWpjD;PiPJI*d3}CEkV{(AgyoKulnBHrg8F>lKyFUP
z5k+OFB#@wnNEvr8-92K61}?^?bDr4}Fv&u)9~$hwi^MPsw~iocYQlmVHI`B*-ZQSo
zs5Dtt;;)v5wuoK`!6_2Sd1@5ddle$bY*iG?Z@1$~yf3GbK6YHHL|+l90RIl)9FBd_
zyK^lwgh@!|gwRiZWSCQA_X2ZIZo%-k25LcVTvOKw9p)ak#hE2G{=7fV*!&T=5n?4{
zI$aDmuQKM=?F6xKn(UB>=K4l*(87eijh<h{V`C_7<cMV?EIXy1{gH&B#kc@`?jDaD
zBk_*Hug8+?rm9taFWqTT7yfy7O9nJ(L-PdmaDMY-1s=IjR)e2grtq=QgDz98MfKTU
z{-3RjL+_1yUDl#HpK4pncpR|d5scJf>A-2!DBY9an@}k(Yx~I2VYNDwZoqxW`<*os
zm(ng-1D}0Res|3T)N>WG{9f73Q7wHJ<VykKx{?inAdW!>kgegYK!Xs|!J8t|wOm}z
zDk3<+Tuivv{<3txN{-Qi-DonQVC_xS+b#;zl9!wSYE5=KCV}X9RL3B`YW+$_x3Hf1
zY(~&*rb^{YOhxSK-{rEY*^n-4o>;0~(0+&;u=+>X^5oxK==x@FG!C!Ty)$BHOZUcD
zQX~ODG5FN$0eXXIUk^JlGwVo>uKU~Th^*gvlx4e^mQI5TMJ@n?-4Rm3u%fznUP{Cx
zS+ynL{x9+1+VEwO)I<eLjSEsnDjGcHp6l+g^q;makG8>6a{%-3EI)SWSdA;69u?MZ
zzENZ|=N!m`F>TvK+tQ#0IZM7dDvh7_v1&!U(^M(=f5d5{gC@X_VLdJ@!OIY^HG#t)
z)?!oxEb#>B!v$#!630iP<xXB2U8>VuF)PA(7whS12bsW@ld_la$2g%51Nx<9CHOb@
z^B1^2JJRB-6N|8df4(&p+s1o`#3>H9n&L<B!=-QEz)$bb=<EI_=tmJ=UO6=+p1iU>
z_qA|3k(1N^8o+NsCF)H!J%Tq&wFtR38_(kRy5xuou-W@uojD2vl=Aw~v0iy_7xWYV
zUf=CMssdCE4*hNi3&~zvB|fKw7g<w+(9NCd($<($&sJh31WF+6S*bhZ?siQ9nIwn@
z!3lg%T7U48=rvT6Z+==Pi=Tk7so9n`&LcuzLWwLSG;J!;V!;n&ZU-7;<-kn$ENsAU
z+ASGW0lhQwDeoj+MOGrkJ0yrGnB;#`9BrVzWrmm>%VM@d987dbz2FJEX{^&KhJH)f
zyE@epk6*4B;Ar5rlZ=3Io5*KMjgD0i?%^S_BuR`^hEy(mF2o=ey<xOkU`{y&*o}G=
zj~kNNuXi!V?a+I`=4$#6en_xq$zn1t173Sm>9?e+B@h+*`)^qGTur%N75-QU3-Lj7
zSFU#<@gzGjv!hs)5c3rxY**Cg!Wk4|VM@6RT3ckM-#b+?Sso^vG&}iJfoW^aBmC}O
z)8K{GeiC}0M2E1zorI_N$tYzEBt8X3FI3+IGvQfY9`D}s=zs(Dc3)U%aId!byUyYq
zW4pEccfs~qfe@n9Jdr*zCV<XBs&R$iQ7onYE9pZy7*uCf3i^^Cdl`NGFfFz33j%|4
zYdO%vOktg1noJmp-Z<|u_FPdKkv?0sg9v6xEAXz;{_rLy4^+BFbMYTBk^$4pf`P=N
zsyP}lN7+4}3e5Pq8&Okj)M-pdmMWY25pANJS4^e@R=Gf}z7%m4@ZN}*OVhQT(I5T~
zf1f=CD(pPUJLyq-A8Wuj7)Vuj@0Z-%7n>0sW-IG=cw!_4cKa|$N=_;pfgseG+&<xV
zv?NptI9hNVnJ3**SO!w?9dmKadn;R8PV4aSUoXI#ht=IY&|}79v5;y_9AKVx055rd
z4LxQd!7Sow0!?NAG-@{%&RUi@|GnILF)T|{CaJ<6O0SW2QnHJ#Ypg6TwfrYFepZ;_
z=aP?OKOk7A_F=T~=)c8XRT@WPO%u1&8O$qH>vLjnap&P-JD{@*IxmtQIN7}HE#$yz
z!0$PpIGg-!z?tFviLlrUepKSr>6RSe-E%|CrzD&FoBHy+^{Ac&=dLH9KvODbawleK
z59?rb@w$~b+l)rdGM_DR=?g-o-z{`}dJU6GzDlGX8-fHgMk$FvWRmIuAb83t3~&0B
zJk9C@c^|J>sUG~p1{V8?&_L5w6z8X$#4ZnW0k1o?36@D2`6qXZzH6jiC&?vx*W91f
zY%%N<?*X@iQ=Y7&#RaM}JMmjG$7QYvbwFI4<kR+J8{vjShBngepQ<r5Sq2GqbZRmT
z8!c&yJNaGgS_p@R^e21nQ;>>c67)fcjcqTnLndw)ev#Q&zWw_H1p=oKIUVgFzaW>9
z5624V=$+pd?gb}@SQOG_ipp2cVdHzh`LRi;gCfA7mDhLg4uH4o)r(mw#l>7qq$t9~
zC?~U}-8_`}WUMCC(P&|{QL{Jr8Z$EM?tipsqQ`UfqSuBe37GY&Udw3KT&G_zpZ+OP
zUMmfkfov<_TO9WSrj8TNmS@^d60K|b*Z(r;h<{8R_`Lxl;=y+O<6-`Tao~_5d=~py
zbsAr%o@A3ZE{*VXOGQC3mi6uSndjIGJb%OJv6V8P){4Fs7BSC8xZN<*b{RoYUxc;h
z$^0WUPZ6)N2Jd8Vqw~Fv5QaQ&Awcqf<1bZo&Iyk|Q%JL;cj<Hq^q4Gnv@6eYN54Rj
zGV={|0eNqV;<jVae4+wiyn7V+TI(9yTzLX+rIx%OI&I#0R$zAq;`RJPdQD{Ph{=ka
z6M3dJoTt54b@mkai-Fr7o~o%|A<Ehqa&wpXF-Htv;1F@1&n@U=YD7@X=L*nDN4CCs
zAdii*AdFQKK>jYN<~Ne4RZml?QSF5<MYDz39K<HYi#AiNV#N+$2D|aIu}ckQUnw~A
zM$iYlD6jR(YQY?YJu0dC`=k8B9EkJV(I@|$n4{xRzG2r&rmz*u)>Uh68@<F+3a&66
z?u*|7dvl9^dh<x?SxI19Flmo$aQzjC98UJ(hTc_3#^uP`XQd0N40&R2)I!py#P(Cc
zij4gLS~QBI6*L^kAa8gsF!rl8gxWJE1VQI=Jg^-W$jsW;#1=Z+(A}4hP_ZD4grI!E
z?Mi$aK7O-TyK@=c_hd$38>zhopH!wNQM7g5D(K!kdXri7U171x|GAt7qrXstTDCYd
zN*vDUzp$-zJ?AgY;(KFlZCFB$`Df;soG<$w-{B(Yl7gRlPq-`6hTJ+1Inllk*Cc!O
z$+)E8#^#t|$VSe@xF6eIi|z=Ia`Fl*28J;wkPiFRQ;~j&LD)zvXtk2aeVBGP)c9L^
zaMS?R>QHIcRsp%%BNCzx*Yr~ZW9XfT8x}#-=rvKMhgW4KcN?hqLuT0<3QlJ;*6Dfm
zVGPt5*X9+*#@VyXO>={9M|A(Lcy(^xm$DAK5Xoabsv5EU=Yow{e{RIRx?I{huM>CD
z9x}yeFQgMR#3gx?K6hj!w`wo97APFo37@_`j>R_ts-rk!-=;_TL9SO?SBv!i@qTqf
z7|IdA;jRM|69RgFaF}SeF(dc@nvkAk`$aQC*yb|IIGAi#EiiH_<WjU;$C1e*11V1n
zyX(rQYgRq2I<BA1*eFRtpePw(skA-80rp{&WO4eQ{U8pBJf18$f0-K>B9pow0f+<&
zSw=yH$LK-@glvB?XLBu84x*!`0&L<aw%AAY3G8<;26x`TV7o$2b7VbBaZZygQfH^&
z79`T+t*zGG^Jq}b8s@Z=V*r#&fJW?dthI(K0Qn}^n58!hHmLz!WCUa6@k=1v1vDV+
zufHWpbY3mf2pf}|5k6j#H9;#Wg~Yru33ZRr#In$68K4Z;N6+i&z_VW@OZTfiim7(>
zLrdvh^Q$a{RM*ZlFBdE&w{K$WMGuIBKh7S?*`0C(p0*iTO-jV(<ijwq&Deh@tNI_S
z4(-I`Fb=F-i-Y2Y)1ZomeerOjKl2!+%eQF}h9j&p5zj^ujshohWLqn=Uk_B({U~mX
zc+jn>M}SkC9u<I@r2M4C_63^70YRcSFpip??|j6qrQ@}sA4sQBppw^(cc&pwGapxw
z6%+F~d~G|6AU(24iQc}DDQGj3PgW$DhLnVyQ>`m%pN43ed7jBvFDNEjSiN($)o(Gh
z6o*LKO7J3kwN`18Yc+!g{KF^NndZ9BZq3;K8dr5-fEh1NLT>A3sc-woH)mi!g|CSU
zn7eU~^h<lqERMkgMWe$!#N*Tb>NbYMzhs%-1*_iLG!o00WJq*9e9(O~!b5hZ56t2h
zGAjh@>x$>m(RFT@MCY;mEKblJG=+amQ=KOOO5rjMr#hwxPks5Br%ocr`?;_yO}*t}
zSVJFsySVCpM#NF)=dys&WJTFiEucp9-;dUNXAk4PD(Ukku_UvhDM|DABU7(J8BLSD
zRt<3s!rangI!Ho3w+s5)`BEggfEH_WyPDM5BNb@l={IS(E93Kx9f^|juh-b}4jU(R
zJfkuqRb-xl1G(3L0d&Oc#YBG%E(z;ENy@+V$+HT5RHU5XgnCPeNeS+e8W65y04mQM
zm<<eS;HdLc`?5CqYxrzg>C6a%K_0DyO9@5P75;zd*SL%tC%0s_4E&Rq1m$|XN?~s~
z|2@hL`Ko~mB}DLd+x$F3xkio%2Me#|TTGhqnoERrRzhL!eGRSnBAP*Iy+?P9OKEx{
zax%992gKg+<!-yjN?|rDnn9~1_>4>ItxR?Rz6lT)Mq8u4brn_#VP;|>akG>Ix9+Cm
zvjEFu$g6NMLX-(0joOl+P!vog-Sv7Q7Ec&o#axhpPZMN+v>HoH3XV=kvqtUr&q#X*
z9Q1MNROF5<WZcy=X%$1ey?{L1dLeq8)i1B%{^&m?lBUY2y7KhzdDJgGrivG-Phw1B
z{;5(G^$uGV;B6xjM5}d<i6jjU!`KG2(?cd}a<bDutVM_~{cvg*;{8QVdJ8nZGW+UW
z$R{3L$Mnyaf=qc{=&J97WyCJA4z`B_JVg&<(+^~RyUtERzIPD0BdAD*3d)9dFO@KS
zAg+J2ghB#$N!(4!OK?UYk^xMR-97i!wYl4E+kmXci&TA(v@UcO$&=W1_UyK$nnDQV
zUWs9P`ThR$Zi46asct}tcf?%y;KCc7ov=M)^S%m;x=c;fYjhU~m;E`X0-x4mJs`27
za{TFeHk89rUIzb`cl#+^f%PTHA`J3dr(v!(Ov}@stbim^HvrHft7d5g=~qpo_4w|8
zWH5}E*tk>F(;+ksv>gcp@HS=t(bgp!FPaDJzeE8IxWW1u(4uZFp}^z^cY7z8YZ8RC
ziZhm_eC3wWTkCne8yPaIVblz?WLS0ix+4-TZnf<;LY-ci=ip&lv~$u-C>3;^XF=MX
z3MwR1cZgEk;i590(zv_Z-<0cr5I($iSm<l$`*{&c#U=QfPP9*Nso$2QD~*lg%Z{~R
z^#W!^t)0EJSA~DnyB82K#=j!@P4fi#YHoLt<Bs#?2FN$7SP^=0hnf6nu7!IIT>xDw
zmlOkC)GaiqwR{Ptu-0Zge=Bmgu!4zrZ?Sr6B3}0E^R(pWlOx<h13hJ<w!h2?o+MUl
zI6~(b5y^Tcd9{;f@8**YkIKDJ>R%~vU`^CQC`i&`oHQv1LbZ*C%oxWq8CeWLc%q2D
zCeQxh2!d`2x|#J+DB6}PjGmUL#{m}}t<DANz)j20HILZIxr89#DDZFCz`}#|?_2Z4
zs{=-Z3VP5zpIP$ST_`Us1J&Svy{r1<Y)+E%{;z$4e2{NE95MS~MukRU0>|ASH^5^6
zOWH^HCmFUbsn`#gDImLmQ}@i))Ti#qz?pV84{BV$Ouh8c3H5qlJ1spZ2_B!5+3cDc
zajvMciz=ELEcvr6y-R5_hcw7_Wrv468(v$Z$0#kRyF3q6Gb#0_XUu*XnH<&#{vJAb
zTvA7}biU}_X0M9HCrEEXlzc;Ay@6d+M(=!f`F$0EA`AKLrP;b@amo7U+REW1=rZ++
zY0t=|F`hMN*=+dn5|1()opASJ($Zus{Z~elT}4$SR<*5q*K5hN9~oSU_|$?$c_=M$
zteRYJViHVHAqyO#Hr3Sgv{?PlcjtaAM9(#t?Ga7RNyS0X%p&WM^7DICG-)bhy<VuI
z6~|*cNa)uoI3D{Ss-Cuq7myncrt@$pm^Ur*<?6EO$?sjT>6*e*{%ffbnp=hnkyqCf
zM2($DILxAbj3!GT!;by<AQrA?vS%ND;|W-LP$U~}IUSc$kU**t|3o-oxe`3=(M60R
z>HUC$MXpLdl1J2nD}=>7`-UzayBwgSu&%=flU|HWwF_5|vd=ESK?_n$eB47yG~CBL
zPN1KtrD|0+jeFpnx^vz^RVqR#LK&)*BCCHcixUjACH1||&nVcy^R)f1rpGhuuLqO?
zu#gcu=ao9dm@&?GtItAI935N8Dr-9s!_E5VWesZa$JyG&gq2kgi9Z*N+2D?^?(`uh
zUuZBQe~m^+sl2!YT0=&X#7~b1M~XZ~n1_*Pr1M@s;O>o`(iNs<Y17m(eTdqXuTP9+
z?9;7PTTU4f#Ma4-pR#$c$B~U&%|I=8Elgbw`)Oi;sjeC8wrbE1<C@%GOjzj$CTLU3
zL8|sFR;LlAhU0E9P682#&Px~k@8qo&TXiR@$e|H4l5U43^bfFz@`=+2Q<E^`pCv!r
z8rg#4PbLAfqD+LaL+Hkg25H>Y{)sJHgzjSejQ|FWH51GqrW{`coE|UJ6CezCjJ{p`
z#4D1{IDyjHu3JU`*3L8jKPsmc06TTVIh!S4)-TZ6jOl8@>S1&q0|-%p+a-}?^o6AU
zK2dC;A)F<5GD(*o4oy(vmC@;VDSF|j17pY16yPA@C$?$?Z!Qz3RV@F?t|Z5KM3PfB
zt*+BgY@2^!`C`B+PRjEsvIkSy`|R2*#2MK8e%~$~J$tR&KOur?X57nF+*ux3cM_mC
zndtvebsy+a8O)%7w1m#re%FL=cl@1l$Gcg}87PC&wdm6@?8B%YnO+F$`p>aKx6}1<
zWc!;qoBj^KYq#s=EY>OvfY(9#4NdsIU2WP=uHzZQdF{gpcFf+gx^c~SCJX7cp1{n1
z89b2^jn0n_L6nyO7~TMb(e=kdTCG|{WDW3G4}T75I`&Xek-ySm%9UD^+ehQbahcWw
zT3p0reL&{`R%tlc!pa{~VLTYr8Km!G(Ox@ruPAeO<37sKE+{^)@?rXy2TBfNi9^Of
z-!QrIYuZSx@Ow9|FQ(QKIQ*RxBEI_!#2<0!I7Z0Z_AZ(L4x?4So8VA{6{j|2-m=R6
z6S%mY9~ZXG`N8sPdEku|T0*H!4ssa<7DxYV*+;u(%o_THqOoaOhm$3Dm!pyYQA<yI
zXZ6nRT(mT{6fPmOBU^UA9KWp?67=W0WJH)A!0a^XB$!GIFCZjJ{FjYdL}Y1qD85uO
z=ClB#L<HCOfHWQc?=QT+>FCCRwLNC|$yYJ;yR@X3{NG36*g~02TmEX~C}W+}NNC>I
zU^~edtx1PqE8gEWgu<`9tlY}8J7!ZpSAvIEmTZv_rzWnulO4e0<`?!mL~e9Vk&MJl
zGn=Cob+#8-J^el7jJ*@SNnZdu0fmER1?1HQD-d8h499n0>b{q-7gtx32Oio9(E9(t
zw`Tp{_|}Xp4DA0S-<pwyjh*@b$hWp>H*zM^Mxw<jGU|F7Zry0G*>X4h|CH+m!Le4Z
z)}*i5>70%$tt+iFIfJ!IovzC@LSm#QV6H0n^(>|Z#sy|a;N+2DmUMLuOu*=88vgO|
zU|bp49NHUyqR`pWZ2%V~WdFtrUjGQp{^{}Y=O7pYYqPrpa|_!8n0%JfPd9{oQzLUj
z3#$tV`KG1@rxuo0fcwtQPN;USj(P6R`=vjnVrxq%`&I_fbdAk)px_f!7SfYpz~n>6
zN&p9?{EDIiSb>3&@zF&L0wW7+C}*Y;p!5w+AnV^Ffan{4veduBW05~QAOSTc@iV{h
z49?9BAY)M#kdfmPWq`s#7^*=4;e1E|vBNya{3hf#esd)w;x|8;|M1qoc_v2h{8Q(C
ztACBLw|@=iehiF&(ljwSfMTVmsr~pce~U6m=BD8Ke{$=?BfotWeh{2LL+}D0aiPip
z$VyFq^K@)%j-jok0RIfQwY<>Hl*Ik3b0R8;5eSUUjDQea+}J=f(=)%(ZOpAKjvoG^
z{`w#!&4&D*5kNBkiU^MUgi>V<Q*TSjpi*c5By3I1pW>ANVym)#&G`Q;pYtyLGy#7%
zsv>abN00v4zRlA8+Kl8=(~^)AN<a26zqx6lSX!AIKr(@obN&>T=NFFs0{qHQVQBxH
zg#St&_p*Q(e)#oP=QkD>ar0*+{_2PgzSq9$O1}Q&g@jyhb)BcBvjMo!$Xo%Cxv`o3
zqatR`{)E@oM-~=P&cE$8{xVMbe*5MomX~*y-^Vr^a~S37RWz;4L+Z>Iv`I;3wxa{6
zW|)xAC+~KxY*3~Qppk!uq_(+Hx4qfj12+JOdv~I6C%VuKhd7kpk18(A3+;DIw$qRj
z|2ge#*07HRu7%q7P(!fdXc=*aUxjO(MZdHNfL>axR&V#UD8iX|QyfY!fSu$3$B%D4
zR6)ZFI8f<8ElbzXyD#}<VjXiv#Sc;!>Ct;x=^K0qcHOTTb9|zOpaQ;z8EAfd!H4Fg
z`?047Sy&5>`s0W8#aeR^;nrWVHonIND=%AR_;x!fhG06T9$QprfIY375lzgkl?Pk6
z-9`tcRb-NqSLc9Yne3M|4vvp3;sEIpU^|=|!8Nxs8*yz_YB$JyYC`+8gxqJC!L2At
zu3->oWKHnzI?oPih_<_{{X1+-3+K*x%|>pk+;ly_sD}SxdEVq>@71u?iXBcK#PMeM
zPH4G9ZZe>czOPd(pd(Qua1fMFe3qqv<{$YgwcAgsRYr>61}hbc;3iXW8lE2ce7$5X
z5rSanC9?HbFCYZhQSf`{*QP~?$*T}M>l@4ro?LA-g6m5E^>+TJJGOu!qMyvzYmty|
zG^qD&`u`X^hu~b8uuaFdZQHhOTPL<{+qUz@wr$%d$%$>xH&ZpU`WLh8>Rm6os{6UG
z`&xLpV<_}uUS1Zii8O>6<PzeXv9U1!ThC@yL*D>})AAKKLH2jrd#5>Mfz?w>sT9LS
zZMwxFj;gvC0@&r|Fw_fK9uvZ&9Jk2Hx$7h}m|y+B!5Y!ngF<}7Qhbdd4ANzd+>T2S
zO!CrdKQ(HHH4y$p*0aUPfDW@aBLb~C#*e2c{x6LH(w~^1ucM^=E84xp(IJ89M7!S3
z9{^1C5TyCxXXh%p(!-8C3R&1$YKJ54JhVN(mFYOL8<K#fG|@yDXCpuVnLIbHUubf<
z4g4vYg~O~33xrBg@Tsak+hnfuaN922sFg|D7G@kc<Y*YuY9;(RX01SWFkpK(F2aXL
zslFsVqJTnh<u9#FE_#?{hB)3Nf|wNxw0U5$tO~bm4YDk*d7!(7CqC(8(401vJLdh5
z)1Vv+QT8-)jaO~w@)ML?D&06S*o5WZidudBLTqcfy2P#CH&nze$zL++cPN#B3N%6_
zg8X7tjI**aWP8q8ZEfH;rLOh85DBMION>+HuT_(mtwGfb6<|UHmJi-`E(@Cm4cG@s
z<rugttZBs<%KR#Snw!%?^=F4lj~GjXbN1hf6Q)G@#0HjT6|$q2O4R2Rhgjh^mLyG*
zo~MPMM5sf0b82;!ZBAnW&I=1S>!N=VSDhD_z8n*y#5h<C3)}h%)e=vC@Ib_K2a;DG
zWjP<V5<CHyS}BhF`8`EjC?rKj<Wx+r9}}O+;Tium=eK;8C*D}m)BYmLY%F=Qm-8cL
zgfjD}iw??7f7y=ZvL9_z#5<cj3AUR_%*^0%nCPGqCJ<W|@{3ZP!$Pf}Bq`BdPXGhQ
zfOnS7_v$_e=#DLfVXhio#D?V84f4|}KZ!xVt%G_5y5ydQ#K#hn9h5safng_;i;<1f
zt&DLii@BoA?$GX`TF>FQ3{f$<T9nkD-8(s@ghr=vUf4nsEOx6}uKzuW$06<@KA|-o
z8&pJPR?bs1<1>z4SH_*TnUCR6V)c`8^<Nh7B?w~h#d{Mg;%cR_q+68nLfIe+RP4fP
zQRxfG9Q5<m+231R8cm6+eRLgj1I^r=OaR4k;9^+_p*7yVN5D!WyaV(LEals4I|ij_
z1;O}%OV(pv*>;~KP?ZdqVw$t`n~+LD*JVkkk>gfa=f(C+IxRwnPD=GR3p7dQM0)}v
zNRnEAQ3gfqvJL#x$rnfNGqi|zTUAYife~@ja4QC8F5j#U(W*;9*ennq<cy%wcIu!b
z9i9Vz_^Q3BDgfYRgUC8l$({6Gj~cl7idfLfUEi~LNY;vWgFs8c&soC~<7#471I4h2
zpGq(4<e2sI!!-VOZnooE8|;PtuaXOQgF3@$r!HGN{Mdi4Ab1^5x>v?uoynI6fr&+V
zlYwP!o!JwZ9ge186Y{qsDDr}AqrN&*Bt9*#LP2#Mq1-sLIgp$KYZYAgHejxg1A)gC
zutf`4ku7|^rq;skhG138fShSN#N6Xm3qmZ6Q`+Fo*n@*#6d)1>1z8gkKt#{H-8;7F
z%`F8C|I~}#CJ7&S)lTY!726<n@NdFq>xJJC52U$wG1zA$*iX?bcWyffAY&|F4nq#l
z8|hDWEO0b#pr4}odE;Fq>!RN-<^UDgl6CDm#@tB)xU*HL_l~75LcMArR~Pw%)Sc2`
z%VkPxT+Q6vS+%@@hh+Vw1TYjvE+}FQQSOYex{7P;f7R<2Oh)wCpB4`@1*TgG3$|o#
zQsyq{jPHErriz#TPA#ug!@`5WaE{zQ_S%xCy*7CTAv_W5SXWEDt3CRjj!fvL(f@mG
zATj9&)$WImjg;pI2!}|JfNq{+*>Rm3)WpAJV!vi_GRq5sJ`T@9{keJ3vr>;q8qS&{
zVC%IX?80@J{zjCwW<zUN;YB-q=8aocIu`ZCRt9E2W1}jwQ%iGap=YT|EeZP5(4GZt
z6+O-RELARa<L`6Zl|rO!T~J27)V}YYHsiJEbLJ1E-oWloKChTR5ym+2kjEXh<bj@S
zL0%h$mHII5KbnLk*3J3~370N+@%DZFF7d7?X++;#AUu*7C7z~Km+!Hy=UEhAvrMLX
zMyB9;N@gVaHLvH7n4}Zmh0n!PK636gT{o(P?jZLQ?d=DrL3AE%IKin(D9zSWd92{q
zYBCqpYS^T%)Wk^X&LDyNiqW<XwsHuVi+2k!DPFg27)$1}5E1H``w<w7UHjQ#gu~>i
zosixTt>hV6!r#b1JPNNvl4FvnV!YO0Coc?*f;=vNBOKXonk{%T-X@`AI<<t>_dS}u
z-nIMZ{E|Z0G5k<1V4a2>rY|YJStQQv5Vtzu0561aBo$_V)~HUmuAH&Mh%HR;&zp}(
zW4)RwMO~?oTNJ~K1kLGeu$?FG*s<#Tle#H+AN)K8nK`>VQTmPQ>s&|2riGc({=U^c
zhs>c2nG)hqajg0%jtS;h2fZL!Y&ZyaSX{nRM149rrF-62h^GCcU6Sn@R)|SS4N9>p
z*@EitA8Phx&SqDcc|LbyQr2;+bei~E2CejXoAXJyDvCBD;Xbk2M6i~-<E;FR9WL}>
zJjc+f?T?j#rsb+p2N_(p;^~j*DW@SKW(xtX5l{!I-HO5Atx!(Bt&-y3tpbs8=|)Wr
zboD}4lS1qFe8<gvuFd({>B|L^AVqNOdCO83aeyZvz7O*+-Jh&FH(v(cL~y8HE#)G>
z(xlaz%~;TKUAK<vOM)CK&V5seorlEhsOc&U(WDNZ#(g;YZ7n*{DW{z<1-GXn8{ZT?
z609S7WL{9{m%K!W;JLx7a88$^T3Ic16*6Ik81s*gr(9+$6{tN%%?lh!now(Rf8AU<
ziUZix&#2zJD)z&nrZ^*M5#lXuMSet@$ICznF2ACxX-FCDjfQO)+AKmZ0F+c_$5Oj0
z8Wi;94UqvfuXOS9@1gY6f;sy{D|g>==FMlbv*UyvhH$aa>{J|??)?sRSH=7OkR<d(
z5MwIZ+Mc!~KVu?$8+STl#l-O=J_WvbC%+nF;{JC|GF(%i+~9G+o~|1#ynDuex3y{w
z!yF5LPkEgyzey@FqCcpgJpH!VfdN&f9xIdc9*HbF#?&j;?o`h_Mw0;vc!T5n2{DW7
z2}59_*amIGTjYRL8jIXiSu%G&A7e!Hg8t3^xCFh<1~az~Yj(YDkNj$`9<6YsZkcVk
z=Sx@ka4J1Rwo}^|ae^osscl+$i{lHT(~RqU&(4e3(~^nj$*KL?(ri=y_tp0W*fRDV
z#O=Eig^C1LeA3Ti9lNBRSSu<p5jIKs{;&XMV%N99erLSM(Zf3k`9J1!eqE4`Lwg!J
zzXYxE<J`snawpmeE>32rm!GR{i|n5^R4d*?ImlK*O-W$8mu%<RO*c4s)dz5$_y1iI
zaqj;$;F^XFE7N57GF;0mbk|F)54*tk1mO&|shH-P(llAEY_BoZuwaeznt-yk;Bd~M
zkXsrGNbKp(#v@xM2FW=f->3AOD@*`odPMiv_+|Gc!p(Zd8h0G6S1;|(6Tbil=-}Bq
zUEQ%*NSx9kcJtz8uLgx=iRKG|35B@WYleZ>;1|8wb+)mYFAJR1tJvB)n-(EIiC+yr
z%)sVo^NhFew6@t6L$DgcM7c?B4CwN-!b_z^muu_tiemhUZL`{>_38S;hly_OmFbd$
zD}oz|jOT@Tt68PFQ(O=!NxQa}*%Zy?|3iCI6(YVc#Ti5>?!_37&(6mB?E8AOZOciJ
z`Se(GE}<r3c8s)!meA>;Hv>jW_9SA9lk-K6m`hhV`j^+r4ob0A8HT$`W}nu>vc)}W
ztlyRaJ8_rGUQ&0P;NE3jHNzo02I!_zUE>X6lliX?a0VX6@s_>*hDeNmtjQKg?EzUN
zN$vExIf9dl(#}6~Xj5Tl+t{`#Rpb75chVS#JVIPZ?Ap4~1XRvbtth)v7Prhh2c}Y<
zuA9a6V}i|7`FGQN2rv$1k%sUDQ;lYr@fqJDUUpsRunujI5Z*T>#;uTk^diIQ6Ks>C
z^rHe=h(ptOrql1my&wUXLjoRU0>MEmS&vB(AuHvB0SrDd-;HTJ52i<BM1=%og*jQ_
zW<NX@mgCTBYsj2W0mRNl;A#{Oshi+FJw5@u#v;jNYQyvzUQCr{VSN1>_oW*4Rb{IE
z0#^V2_}dWmD40aWyP7nhHb(7AfAw{D7vBtelfD`wEYI;2)1ar@)`Csl+A*FlMf2Ag
zgr8+qIZ!7<wz6^T897SY5;o}NVS1M^Fc;GAt|BuT2+1NzKj=%HivJSbFWin_3A=NA
zdvk_PH&UNn<(-6G1neHA5nZ7fasT8VQNRaRt$kw?D%A<P@mj~Wj3LnOhMn(vUXYaX
z9aKxPhJ3%eVk?x`f2?wq8gFip7b#?@^A=+iIvGSgbd%WUntZCg!i-@vT`0KR&{Y>v
zk&8%|_P)?+L&W5%qV2{>SX>b_JvN3<Jr9g$Af5W~(2RnCY70E?NFmfxJu46v%5b_~
zC{gHp9Y5Sa1%er78Y2g^K*P{nE-d6hER*h6LBNv~SI#9E_0iZE8ra9ru>`fJ6U|l%
zLP%rC`EytQGrK+3f}<Th&PL5t*NR!r0`C?XvTp+xd|Z5`Q{i)a-!xoMUQoBbZYhdq
zCHs`+gR4Yo8@U9w@?>=!XtXt-jM>7=BLxkKXj?^y+a+aoE4$uPs<EGvgCP)yj@7*O
zqMVJf2yushKxn&_7j?gkVfK3(`q1UzH8{m(7D-_>Ccl}V<~rF6*Xg#(zHo#_ip8to
zyfp1XpOiSi=-c^+Fk6Ie<A+&;DBrPQMNE!9y#jm`YmSs0Sehe5iY^7xc~(yd{IAl?
z*!3n3M-<~lU18-&9Ziv=>1i*tULgK-D*N9OQE~KfVcw97E9Fd$Zeo+MBq$gE6DtA|
ztMyn*1ZXnA5vicb7X@*s^$Rsb?(8Z5J?mF12TUN4jQYbrdCw`+-9Kz-T6(u{oy)N%
zNoSO&wi;QH?-q!7p1Fm6O>ku~3G-I<CNRzPG~kG{C2UJTYm!KjcA%z?%~h^Ht<gNa
zWnQIxmd0PJKbz$vKXv?}5)%NuuXe-|Azs>g<vu|v^lNY-xc!h~a$h8s9dE*Pdqo;9
ztHxrLH>pY20zIAQ#MiYcA!*=dKk9{kZL7!HQ!o@WJc3g5s5`#b*HvGhe!-KB$X@8m
z8SbOk@u?*h(;}>tsS%zM`?#B|p<F~whFLr>nj;0SGx!En9nO|yrO$8%{iaE7hiIZO
z={ti@Z)Kg+@S8=8891h89lPd{QJQx2#)hiZU$hqh)%#NlB7V1LxZ?MMK+ASaafi#x
zZOnP66OR;!59KnGa>E5|7PAJXE>6TzjzJ~Zq6C=N2fx>m3*|9<b9`9)@bcBR!>h!z
zx$}=tc^Lmq#p0y0uk#yAzp+PDPr@#fG8)yOi$I1D7~yPOP%=kB$(DA+GcL?)#ti~9
zAd1`L(EF;sOr;+y=6pNKA^7|ZA#!$lg_sGM?i}7&p650IndNw1BJuD{t!>HB?m;Z~
z?C$`(oMzBv>{5r8*N7pR$y0Ez*@Baw`jHj5oRjU1Q*Xg^F)`f8YAcA!3-Q^#1IB3T
zVXoE87^$M~G9@~P>!~8v33{|-_EmVwHQ#cu#SkGmryYi_qu?qLV?012OEO!y?)W(X
z>z9nY1WZ8M_v{Ax*Dbpz$jEe(nSGJU@XyZGF5cTb<9&}<JaE2JsfS3OHbLP_gl>rf
zAsTU(m<9i_iR^Tp!9dOr06<G){cSG2dF$$f^MN6voffuOWFe?#Xl%4RHr(DDhjW?o
zpe`T&9v8D#U+;YRx&OPMiq=&+0P{yCs3qetFs7nj{RBQ<2_ofcq0|MSyz%OM*A}M|
zhap=-H#{z4P-yy)J$}GHD&I?Oh<Ak)j_!@x6uIEuj{wQzqkx@tQ;wNMF#zJ$%R=fd
zuAXY_oB)J=EkM566M;Wv5DJcj@EhnQF$w)w<R`0=cJMAwZa)dz+^mLQqzxD{5`Q=#
zg$LH5Zwq0)Fq3TmN3CR{?e0ZN#{d4;8D4<74*z3J;Lv0IMbtKIX7C2Qy@v7NR~2?m
z5APVkT9JtTVQ`2?Hk>WEWPT6A6ef*m<DgnEzo&DGFE8{X(tcq^a^;*lTFx;88Cg4;
zDlAr_hmW+Ra|1(ZSq4D|4<*k)XuW+LS%=?F&4bRu#Lsd((wyJJVXz0bzYE_Qir{J5
zhB!r5+DE(U!iIM4K7?xf@J%_8Rj&xYCuO<^7~!WKjLh13>ei=B$CA+Qob4v2K=)et
zt5ZuWDSs!;|BbbS37sRaE*f`OB@5r^baT9NicvQqIwI9#Q|)X~ICHkkk^;-+KhgSx
zR9!b)LbSr@jzJJl%fE1__h>PQzSVO(sl*GTYBXwIg+`o}cM0M(lk-r9MaM=GWtA2<
z<`n-E-MD;*4nElb>?2M(lepzW`rG|HoBD^W>X?>d`9%~v()gNlIiSL%d~uP9Lblyr
zWl3!jr4TT;g0}~^oiua9UAu1+WEq-f)u|=a#-}xqFlb$T7D9W1i4m65qgrIP8a|gK
zj7UvMQXsc?rxG5INbQ;N6uFB1xR6op0euz_w{}MpqOk++8sp{N%oM+)%F3d7VUl)p
z4)I|d#8~jI?rg^o);N?Ll3#3Kk?vRP0@*vu`z_t7{kK4B|7Z&?QOyXL!U@PLcD6eT
zl$6GP0@dAlr0zKXatFvrFQ<eKJTHC!<$rck%u+8BTC3wD<3%yUVo(hxkS4O<{B(Im
zmpI;<=*lH-*OX+ruqeV+FD+pIBRzcm>gPm4&M4x`{E$iyPZqd*unWnveaZST7wMEa
z`hCG~L~>6@&PEo1uZXK;N%kYslR!WAuj&tGo>2HgkJrw;9=1Z1(LYWc1jm#9WU2U*
zhZq5<OmnTto#n%2veQ7%WBOmyuX(hS&wIdaEb&u`nB&6)MwSeu-RR4Ud6Un8W;S6I
z5jyq-cBMchE(p%<WlBnVAD87QwF#1ovdZtr!YOT+j%f2*+n<_T<vVANmGF##zkYF<
zRT0bjS<o^q9&?<-f_JI&Nx?5FS2bq^l)<pk{4)&?9NTs~%ML7y{L)XhI?h-YxM30d
z2B;+66j|)H{@U+xTdS=S|K*|1s-!XSn1s<aG%XnaDR?bPTR|Aarzs}z5>(#Fx^p`=
z^ID)iyliA@%hGgX^7-69(o;3T^6*2Dd1SHYm`3Bx^J$Lx)z8VE>KCXRZiO0_Z$}sP
zG1^~N$v&D@8NBZRA_d+QbWhQ>vRnZkPkW<WHM!X1mxCtQIbPY)`Z1wO{eFM}r{<9d
z<(92PY<ET#JR&@OsiE8gTCgCu-8)tKwI(qgX8mS~`SeWQyI8W2Gg(s{gxatUS$J+;
z4+iCQwz49sv11FzmbE*=@I8e3hGBFUCFSLOP6;Z5iPyHD_9Dw5?*gVR`-QC5v(q<=
z-&c<ZG0em?Q+Mw&Ez9SHLxHL_0g%rz@%NfM9-|DQq98;L?nq`TNVBxOigz5(Cz@re
zQH}uB_@m8Co7UU=G<$`fdA$o|Ye#&MK{Q(m$7wd!vK)jEQ@E-oD+&-i{B7siUl2iM
zA#8yj00c?2OCC6dn;~c&>}^iB8H%ea7@y#ka|{EgHajJWFlX!QD0v?>=kT&;l(%-F
z1>YFfhbT=X%-DX{2b)@+%e&PQTPx=~m__DWR79{R**{;;H{W+{Nz;c<Fa~US$zi1Q
z9iBJPaj~$4HyMI%kcs8g8%0iqw;$hTs*lp-d`^Vd5XrsGAR6JzdV7-9f_UN@B|0r(
zwW~EIza#d*ZYlA0Wlo;O^!mlm7iPwLI9JBSUh)F-BFigdsyujC&a`R?etCB)X=Yk~
zTwb^Q^IPEGdbDP`pu(6}eF;Kc;razW-Z<1-Yus)npSlUEbELc+lf?DWHWr&olBfh~
zU721QrvVG)$n_gl;#b7PHheE(9Z01un70{uA@7KgtQm`WM%zq_iTZEtd-H4F*Ukv?
zGqK#M$ZZ)!o*79v3}J2Z3Rq|VoVU4eZ_+m^ghazyWH40zXh6hUKL4@Z7=x`w*X6fr
zxtu;bCSX5g30B0-rhmuzAAq>}g;6p>rRm_A!w9B1D*`i%a-mKFa`BtealM7=IYg`b
zWwZN5#gzTj_;Wp=X@S?J&xw>P=Yxt>uUzFN>fFw{%s@skJ4%vv5w#|F2hfV38(Slo
zaqCPbeSz`|1#fhAYq<9!(wQ-6@ie}(PG@!gYyaFOV9<0CiuhpE($#_=I__KIb(!8G
z)A6*Dcjx*$lnLRqSjSb}jYoHY_+gACTaNIYu^B7ueJy@*K*w=}wz&`>D-S1Q7OZ$n
zm3ZOvUu`ytV+SSl{PZB^cu(z1hS4_-WsmZ<VrJ*uLw*{R%rQ>|mtO%zF$ZSU3bB?f
zivbU7MUD6~_V64#!SnZYYw>(fOkxno@t67^mb(0=8vtsLy@MSgFs26V5WEg<Qk@=$
zdi-@Tyo``8L~CUuq5}51C7UBCT;2Yyx8eJ(*Sx$J>gX+hx!?|dTnC*l5!u4U)aUE=
zGz3Ph%m^19C4@1+JUz{su7C`na>%^_VQW>Ei7&+#dFyG<?TIek&tDKyUs!tp&pzrR
z0jnvb`W*!(txI{q0s^fjGjQH3+<PhCtD<a*xP^MfLq|IVevK`6So_5ZSXLw?{jMKF
z%m&y)-=e{36{0D#D+!fC8K;O)+$I2!`6GGImcXtcU`$yoG$iME`kkR^^oS%hhO?_Q
zBjD$Ck|}nt(cm7-<gnPyx16KQ-+zLQVWnkU5E|*e)uop!@%oAa)153R>wEF1k+?w>
z0rU%XbEeHi{uZBhcc@2oomTC$E{_ccO!&Ra%Mjn-&QL}J)`S%x3X<P72BJN%_qbK@
zu_xnW`PIkRcAP_yFycJe#vlr!7_f#6!8{uw!K}N!P3DBb<6z{D>X}DY7|c$|c)1Tf
zZjH6(l{X8uS*5A$T|{>(GoydUX>lcAt$laIz0anti22Yv`a9#TGZ?fXV#jP?zCNDj
z+lbO%(dZe(w5NC7)YSI2kQmL&;>rXJ;TbOGDcs7^`iXOWlgF{*x(+!NA>J0Kmo`S7
z$;wgu#IV4C4eDbP^|9|n=(_JC)_e<Ny+N#&l#rMzNKZ^mhaa}Piv4iXZ`hsoa3c!u
zU;5y#^LlqIyw|^J7cPG~z&BJC>xRd{Lpae$1A5~px21*DkS`H^Se^c;r<~5lNuuCK
ziX9vNrM?UR9v@soRO~O`hYQI^A7hzPG?t!oE_0A%%G4!8<cg&Kyb-O#bSUY~=pLmr
z;w6ASi<uq84yz!YJP!LT@g3hvme041hh`doL1iWJfZvn9>z(!vJG?Dh&K&m?HhapL
zS-a!M|9~eCCoHc1A^?;rJ`+FKF)zhk&;~c8w&vLuPs5V5ZlKraV^k7)oRB$_Z(wlZ
zs+{0B_{D^rpHFX4x!+O_Xd4%tg-^R*OG)B&%^B?hF-N~as-|#HqyZ(2eCM<kb?U9=
zt4Sg9AAnhO04c&cWLe4$&*J4cO9Pe;p|-sc!KQ$HEYr_J&z(UOT1N8k;~+4_+u{cm
z9esq$X86G3+u7}c%@niDL2!?A@y=}*zdr$KNx*ox!DjH=UV4si7i1g+>UNKh-4%D_
zKTnJ(%sdy6jOYgj5NfJ(v(wn}V9~Q)^5X@+uf4;rf-2e}LR=_Fl5moj1F9?0b6&YJ
zBFP`qi1t3#TJ+7?a4-c8rKE-K3h#$;9FT1;PVFxXzr+I6J{tBp8ajMD*G@ruW_Zj9
zAs%J`79Z7jx0W}Kg+_IyKLrI>&%7m6fjm*=7!u_B^?q}fHFw4JJ_rK763!()JuO4X
zTk2|mfk!hMewe&gXviUpQEOr2ljtOaW<)u5+MwRoJ}as*<N+K10&jrdQ2cHuDk`X#
zpOH`wOP{dFZYM$F=hmvu_*X}Sig+p=#C2FV#XDvs$#Jv^`i;#n`A9p}8%Wl!IPWP<
zd2Ip+JP)+}_N2lgpPS%Cj_#dsB@jHBAh_9GEP;lI#&W*W8QQbEf>@rifUucADibGX
zP;uL-8U&uyJ;+|}Vt(@Rz1wxPP`c?UMfpDFmt=2sD#j=~10a3foz@0%JKo4{713h7
zGmb8Ovy@JBqy5u@$Hl#gbv^y5_7v^Mt7^abD_sfx%^y$$=nqvygJrCVm2RGVt$KR3
z758N7ueQrU1S6<7V0?H!f@iSnCg3Ct6u7{X*8x4o`pYV2PC;(L<E+6op_CxLHC&+-
zbJ!mt9KNyEG-sSsuYR_eZx%5LNQy8PsQ;N-pH9+2tH0>7c42Cbf<?TwV+RSoK7>|`
zejTf7)fK}2gzbwEZ!pzTmgqG~u$1CGldS=~B8i(nuZ;aOCwqiuU^<tj>tS5%KRR-*
zLdMpNYihSsk7A}UOtGDelggbny*h4WIIrAkq~3`5Sl@<*sp2yfOp%Q#K?1(@c{8cd
z<Yi1)@3KgQ_=Lb!&aUX3LL1u+`1+xsjXyutebhq{gg;|<r<9rKVLXyr8*dT>{z>2~
z(4wmO{%GcKcWh`!=rrMm&T<>;zV#?0Q2f%@+@;ZzMbO4WP4M4^IM-KgVEKr`u+aED
zm_?>p&MC1sNk)_D)6PKLVk<vB!ZRWX{eCTf%L*=5ZbEfhR=!Ka4kiD6(WivD(kGiy
z)mOs=Cg@<r7P0=sA5CDey`;8#Vrdv7E#T!Ps{NHUixhWwjn;-X8HL*a7CrTJN65#E
zY5K~gpqo1!k`4+2E^xUU1^z-PqCLTW1$GftSnn!eRLN!%hei;quQKc?C@L3Eh)8BQ
z3ULP^9#|N;XnR+4QU5qOyg$Ez)L7C^$Q52$7G1i9tOOmzX*53C+9v^io!>ug3xV2y
zY5XfaCgeBG=A6`wY``i?C*3mAOPGY8N|4VyeHgZF3n7S2Qc=0xouynQ;n(6rrt*~Z
zy>nsSP|U6=X(;lOjBr<r3j&NT$Q~sYtaHeR0!*Xb77%sWr>ZE}L4k+ta%$=7+MxWI
zf1g0`f!CYm!=xrY>|<Rtb9wQzMEfyTb*S47Hf*dPGN=#x$ZXZ}NTM|^(q81r&-I4<
z$IdrQjQ*%e;#gb^JjMN5l^b7Q2lvW0sh+u{eSp%XB9vgN<Z(ZAKAyI>j@VVWUrTwo
zmnpS^>@_f@J2vY+54KhLoA_xSsC``^N1SG3c0g~2^}oKo>n377Uh&-L>;wXdsM)kN
z1~lekYb7WSp#kMp&vE^v%vx1y)8lZvY4Q<c6aTxAfpEaC9VsRhbAdwG)<)k>>m41;
z=g~53eN;ju;p`JrO2fO7ABxASq<{ecj1Fm5Rfu92%;6&v1JbXG>1&{}#>RopJeDrZ
z1=K4Ht<R3Yz!`a<Ujq5&s}>9E@3Zt$$1+A~Z$NkrK1lFZ(nW#L@<mAG#e(|u@wdoP
z$$6WA+dtTlv?b5oRld&wiFXeMb{AamtW>aEksjHodwdxv=(uPbfw9{%sVXYhJnp>N
zd@u-kjg-$CBlGGddo5#6Gj8k~+qaXq7blYK$5}tax?=AztH;m_)gbA}kE~BqktLv#
zwqe-+Dw6%Zzi{J=bgeJmq3Sc~%ruU?5bTO{|7C~0>?Vk`_WZhht!PTr*{&<|CA;hO
z6k-kOcF5T!T&7gvj6*T^_SS{_4WshagoQ+ECUPVRF>`w!9TCh;x8|~dOgGLkvP#ce
zzu${iu?$@F4G(YfPreFGbH(9+x7g6057!fV{|nsCPT<)=|Ck`UQw`2*u``+aL0Z4V
z@}9-I)=Sz7x=w%)dugl3JZMWA<>*iXhefwl=IF4N)dRIjV^HyWY@cx5r{l=lP16%Z
zcn|hqHBElm<am*afrwNo1?3LumUq>`C}mFVm)Qh@ZE|Gn7)V{gQET}Xcu=>Uu0DwM
zug-2bn_G44TZCMS`kJK<dOo*~ke-*tMyg|2!~&2m&zj`5z?DWc^-7W#ex){yZjQUG
z94gv4zTAuA$b8U7=}BvQS?jgs-Q#3{L?H{}+jo`-IZ426M2w-#Am-D~4uca!zZ2K0
z{nyTichOR|GnSY-@sYBfX+Mp*P77jYlbBhE!wcP6yqMkSK}G#uN7BmjpFt69{RRha
z5$^3)*YmGN*DQN~8L)yrz2av6bSgu{GY(zoRI1_LNpXD5#H{p0IQFOKutgTh--L8q
zSM-4V{bgP>**F}EV5~Jw+>u<!XVfwi)9w?Ya0_j4???RVBx!a!tZcP!Djo_&*xN!4
z;Z@U?tF~+`BA}p_2J)Wmf2bqf_+yu8g#OTC0~NOFG?btFTQ-r0ngTujQZ@`cO!*Bj
z9z{HRL%A_A+uPWlEjM?)MD^dCOd9?tYlv+0n`J>81!fE^C4@nXIhc}enNq{0)FNnl
z@LIYt`qulhhbzH2)2%*#^4kcf_~2vW_s#A7qm<O4w!;c4JA7XOkv4Q#Lcwj94z~g+
z7>whb=iB+0S7{}BvnQI9QAe3|*2fo82yZO%5VLOUa&eGBv4&Yo*aOAVP?yFAC_w>a
zSW=<0vhr!7pR{yld6TK(W^N1y{+V%z526#7p%?xq#@gmW7e*eO$KW{0Xw^a4(Yhbt
zEQC`o`I{nLt}(d*czphP+`}K#s~cg`M+GG#ISaGya)iis#`g}4rG8E!qu7AFGu-##
zQf%*q*fM$&3+sX|klup-O7pMirDS0xxsvBIZ$XkfQYlP;EL<6}UZu;DC~sIP+w+it
zMUI@W204Lyy=n^;#&l5y^O!e66A*_pxVtZaSx#s2s7On+QtR-n{!59OjGcZ#9jc2C
zxIa^4=d~UV$s<qW4&Kka#-!gmz_>71B&H~1k=c=^_?+0iye2JCS36qCni^eOALJA>
z8L8())YghL1B2c_L29;){i%XgCR|ez2o;vX98rN}@;sbp#`;gjNb?b^jJ~=f)iL)E
zj6^d+HnhfTj#6j>JCB|eR3Y75d5GH5U0l<ySA{H$qq-d?PXJy|6U$U!(Bt2Z?m6hZ
z8<@COJ$K1d$U&oa>C1e`q|mc%yM?oeG`XKG7+FjRM-pOMv`>%MuL_73e{2ptto(EH
zKv8HgjA51O={&3hU9*AWNKN!aKg3O9KllfkWRs_8f?g|}^Iun8mIoEWEK|LRmlWH5
z&Wy`wlp@8?kqrH#5{I)hh4y?<Ak2ULoyjch-@65flzA`2Q&c~FC@|nAT@TEWXZHV@
zG@0qthdhq<aacuj2ejRsElQhidoJ7iR+$WCDcmRm(Wd18niDStP2aX)gFg$_#OD;}
zJXfNzwft1poS!!a9_RbN>nd5%bgJhT9gOed#F*$8#Qc!B<1D_fGsH9(ru)4j5?@QH
zikbn@dhygKrakd%p>W}q!nnc>6UtV?Fi{i@4>i-v*<qG2ci|6@ZTUn%{FfS+M_2Ip
zxfx4mob?zi_!Qq@!)x+9w1MYli5ey}oXwVrEyUv(3c#CCOCxI8P$q@vO3=~+L`qus
zm_GP}Gna?JK_@G<cu0s7R$iU}6k&^cVV?d}pUM===Q0=gvZBL7jygM4ZJVVvpkkC-
zo%KtrCkOQ+l;oT|mUAltevb>WSRb7d$(VP7N-Ei7NOH$7^~#S2zaB3rXEo-wmUs<M
zK=d0tG(xF?-$~Y?KBqe|)D75=>-Dddpg*03;<V=jDJ>aXVII`P$LlHQz>N$pO`|xq
z<#!*nRqQcQ;2i{xk|f-l`x-Semx%?tgLGU8I9AS1v?1YcEH1Isk^c2<X*rq!B3-y&
z@QC(sN#_^2viRAVAX}&~taKn##9XRvkuk@12QMgwRA>vCns`9l&^#S_4<3x+E*Zo*
z_CYy37Lv3RPb~>Mx?m&+$k}V<^x^M3|3w{4JIwgGv&#jV7C(S(2O+jrG{W59RyZ=&
zPbs0A4%t}<OZT+Md77!FEY&Q74%wKjuz2fsc`Iz4a<9tdK^)qFJN^C>mLL^!HM1wB
zBkR(nV~k0(3Bu<@p}Pv4G-45?q=a%F`-&5V8uV?GTVm$p!RdnkmKGNL@0dzuu^eqY
zEX4kP5QzF7thjYJwu7ozbvr}i^l)Oy5i<OJ!|n_*RDSr#ZY>0(q)Xz)37w3=XUV~>
zD4<m}MY^hgd~=DKU@2^!dUvLVR#}4q&b<mL2|tf;6i4;cXfh!|u%tW%^xtZ<^CCzm
z=RZq=^NLP!RRD+{&n7qKm$JJO;O|?wk^i9JNOQ9-*&VIX4~N`&O2r-?++Aucb<0?d
z(Bi~ghP{u<+Vyt(&MAm9<d9X{x@%Llpd$S|17D-mb##YY4~h`o-g?Q~pF+*7Tkw!W
zRxJdx?Dw&DQp&FXMXf5cu4Dg83qLL+GpeGtiaXh3DXE-&AIbG@=r(x0nW3+>!_3aw
zQ7E^hy<w@VZ&~^*hB-@RlZ<bACxug5TYQ>hXYQI!-3h%K?k5N@+Ynxd!>*f++Gt6d
z3wKIEn}R4HC?Qh8$e-mf6wRH*uw`#lqluyu>H_tBvE@$;l~zesp=}&BS+M+gbwQr(
z1OoLa9`Lg;BNSvV%$I3ugc&>igx59&B7orrW<`yHP`l)n&f;=~w1X3qYleg9*n;V-
zy-l{~KG{?WE%;7Y<ptv-&IKv{uQ2)WS=3Sezc=|GKe63JF5@Y!w?3BmNt46u?XCNH
zIp7gA?WSIJYw4Iy(EZ~*Hd6dXA>>Z{<p{}SnCVw1xMwmUEolgQ$-6so&qoi*9>uP6
z(Be*7j+=<Y)1d&v=DEBa-rNY{`7v6ms$+os?K~1l7$Q}Njd<+yD2ypQ4q4M|P)|Fh
z`VC?-Jxr`a8lfsz&svkZIV<UwMusO6?Bgr*2OFpMZ3s-Mbp&)0TEA(pdGU4C509S2
zh2{!>KiQ2Zv;q|M<gXmujr+@c`5#<7wf^Mi!91w;prP5#e+W1qR9WEtNPLrauRlF)
zU+$<zPtFF~qqzII$%fgb!UyCFL<KMSy{AF-CN&I0qTkup@Z;mz>XoC7m({V3o!&O?
z`VG}E{Y;Ut!KVc>lWivDX&@KTh?;kNfr3D=Tz>rCN=c1t{IFoLj>h{g;VjWZJOzqG
zDQ^4kW+H`nSo#C`u9U;VOWZi?+l1ZQlY_eUV@U#}3skbtc)!tdnESKfLcH%Cp6^hK
zXzGG3b;~QRA6qf9bFU1JGdF#-B*1GN$b42D>M5nGQGvv#_y!&_5uUx8K1e2x$rieB
zBpC{TFjLg9b%WA70^17%BZ<;-gUPovOKE&~UYzl6|GfWxJ})$aW&pM-Vo7T_l?@u2
z4n_C7t4o}>i?-uZ|MF+4X?g{_9)8(kUj$bdp_NWPV?c-_<fjHnQ%+#0Lc@;wTSYxw
z9hS-4md|dbV)T!bV^n%DDOSys8RZZ6-pQtKAv~>kI-HDkxCG=N5!n#+#7t={83m){
z%=DNAG8CRXCv?ccMWrBLc>O+qxf<#OGrr_78BTxOS{Oh=p!x8pfJGGcVf-ijhiUP;
z)DK<0&fiTOy3b%k<2#~uv{^xRZGW~y0@A)h5`TB4n7YJndVupnk->CH%IzE>6I-S0
zNtMgET;0lFE39hH<6QkF<Ikwr`WTk55Rn~kKEVM*P7@K<tPiu()d3$bya0sB#u|`$
z16A|w!ejsY42wsxI_Vjb&<#KQ5wnyAsxs}fle|dCbz2@eDl~JP$P!lAsko1ADvx25
z8Bcteh`1vNDvM(pMj(*wCag{uyLsN^DINn9rcXT_TXMah8{AcfWdwmUKA#N#k;E-O
z0R~cEEP&-ifbSMq6{)#p4~2cN+8rozT-Qfo`i9_1cQIHh;<SqY7)#}&6i#gTdR%HS
z#6eQ)jb|V!bS=b%mBg}|esHOOx0}_zSU|K(w=fpFJmC(XKFXo<%Oo=ix+Arl6aSJB
z=K>wO1nF;Vs<sw)l&C&N0i+}TNwVlx-O-y99}(+i`#A+8SjwC;-th{3xxpgnaLH)t
zVV-;oa==iay+&F0tolo#3iW`N=kSqd)ieDE0tbW83?R(H&_7?|58BwEuLbFzR4m?b
z;q1~my0oWk&US&>&}~dyxMpNY#vCM?rioa{slRo^IzAs?8-QY?;E&~fPQ-rPPa*88
zTJ%jh?=ulhB6g#J`9a4P_)QjCKnCs8VJ2~Wia=py#Dq-<yt!b6@0|J~`yLF8PsIK3
zh;^R;z#^J6$m8W={<Sw~lpUI9>nUuM;`BM$N2HB;3PGc4SRqGx6WOcPjs;7K_W5(P
zy<nj2-nnsZWY2r|9^WNj;ox`nG@d^^Q1BDT!hhdRKeFnK_k|Wm{o3TM>yQH;#$Cu`
zF;t>_asb5c;om_($Rk&pRielaYA+l~wX%wk0!d;@qzf(iD*(jGja0fhwwV6Zlbpf^
z|1Z`3pNIy>xMPVZ1Bcgn;?ny4^pd84qWgp%(yuhrN3{9R7_tKCEek1XLiG@2>@P{k
z@=(xQxZG?*nUv6ZUUouPAUhP5O=hRqZZaCRzv4_kp$hQQZjA$KzP1(PJJ~MVh6Sen
z=myPMdkkc9MCi0}XPBzYjj~*`azN%%0>zFhJU&SL;5q>rWoU!Ow`><!xRYgjw}0q9
zoA$v72BQfv_;x4XTSqUDk%LZhLqHjopH}!O41syS$T@7=?A#hu9qzT$!PdN3kO~>V
zXAxg&zWg=<6pOy)HJhUmxED?Z^dJ4Q!T^sA{N_x-OBA!)gijRx?L|}^*MQRhqTdv!
z1HAYB36TBs?iv3efBq`vi>RkFk_ovx$GARmlE{KUqtWaDbTVI;Bm0If_BbOz<u}mM
zhku!lXN6h`O}oQpeYWIXs%1o`Z+yQt=x@d|A7B=Yix5Ct2E>L#N@6s4XZIRkqpTJs
zp)9g1a2FWbRgbN7Ixtnp3Ko~(xk?QB7McmlU|Zu2i56W3Jpx=}=RI+^6u%&v(qcV;
zK8PXkXZ>qitDgy4dKbsL_Qf^>4h7Aw4X>W>zlx{XUjN2*?1FJ0$d7I?d65S^@)oW<
zx0L0khR><K16-F~V6+?RGU#7Rfor_dqrT`COFwv7iQ5pVzcaaRr#QfxR2X~iRZrdH
z25sZqvc<RsR6Ez1+8cpZFwJ{a!u6E-%Z{Q+#_`Q7HePIDQ`H&0!+{g7D1XfVf&JoB
z9GeeIIzHp=&%59CF7EUTO7k#`BiJtc?OO=0rVlfx@39SL>%uNQN*WSt!n+=0y%Bej
zE1(w>{ty@|ToYF6atoEygesrz#_0Tsz!5UPzelhz^%tbX;BOi1v8ITq&MLP#{fbJS
z5ugGawhL7FlSuxgE<0>29sD70$i1+&;8>QYR_OopQ8iyoi+Pnk;>aq<c7`u7B~e_A
zOJmGGme8C%JNAXG)-Aj0b?f&Zj2$_PhTO82$GzBNcZecZ>&$6g4vIN6xYhDstyorh
z_LIt&TbkwMP;6`iQk?2&!O9y5HD5y-0xhAmb-ZqewaQdc=)=ew$|TMZz1#LzYI?_&
zXgb_E<Iz#{fJCpd&IMp=9bkXxtMdco$3Ui&({LPQT>(3c1potdp#8d~V;?zg{rgyb
z+U_QxwHRDmO@Da?E{eysv@z8lhn)>`T%_uD{+dAjt?~jgcy-6AztpS=nj|#Qnb~~5
zJ38QFctHfI%u=@J1w|hX>w;Irx>Y@R+C%Hje3z|uB_Exya3|fpcQa9;KxFNMYXcwB
z4`DA$>o_l%&U6R>cDi0-0XImNm?C&sZIJH)KY9_LJXZBNmyp?{603A*U>Md`qS1!6
z)?+)Q!&9J0tvo`jCZ-nW)#gCxmBjBO>FonaOOSF(edWbM<u&58SgO|4@$tL>BJMc+
zH;cU6wYEtz55YWRz#Fi24Xb@J?A+OkJsFMQr+;nQHu{u>uPG8JY6QEq;;OfyLU<j2
z5xy;1lZI@iR4-uwBT#F`Z|9TA-+@lQK8-dUpCM!5mO@eYB*5mvnQp^_5niJkpCg(H
zZgrFCdHYool;dx-n9PW^`KLz(Z7LOX5<i_Hn_5%s_>sauNQGN9$+pH+Z@LXPtd)u}
z-_@Bm@oS#wKNo-i8)qQ>Jw|!TXd)f4G9vu#FA&=Sx5WRzadQ5T949jjqud{R3s*}P
zB4$ob_W!qJC1U!&J4DP}EL=?gpB$%cx2dE3ZW}$`wtN0Yn_UiVTO9hf)_*jof`;vO
zTij=U@2x=li>BeMU>>`ULy6|Bw#$>S1ev|VY4OFm(UjP@*!%>%60&Bx>9LU+7}HYo
zKw?r9R~Al}*4B0uPFJQgD4U!-5KTD(5QRAqD=8@|9JbKN-r(}&(#{aLkd@psBdO5t
z*uu!t+7fi3y@Tnkxs@&Oft!m9s;i58nTPX9#V@7w=nC3_i7^aQa|<(gBo&pl^rS2h
z75VWh5K6gwxk5Mxkhmr(uB>TbO<7}AUNbfNz|0=B(W3*nk*$M;$-7`??)fkHz})Q0
z<O!jXxxz~u5I3_ru@ktLdSE702moqeVGfLl_Fb_nFNXD1aIt($q3u@{pMKyZfE;)D
zH!b!@^TV({WxSu<8;gP>FsiH(c3@&<X9*$vFu*gViM1KB;dk5i@5LSAx?o`L?;vF1
zSA-}{U{uwn;9?yI$HV_YIgx#IAvG~oDFbtRqpDKyg*Nxr7H$NU!`sq1@_+lCemO;^
zM(+p-La%0ijpOh4DP%32WD!+S#bq-BDmhqN-Xn?wHreHVdl{X}DbnlOe&;y9BN77K
zS4Q^c#vXqe-W9QbbrOZOl~q*aGXDkO9vIp)^P2jg%JRAdMo$as;qYXA)gKz#i;GSN
zc+3Y3d{yKC$^9y3u(~=mJpr9DE&VF`F#XIw5}vRB#-L!dz1|g@nwtSHG%~S*WNu_}
z1MAhy2>gPtZ>&yEuFRbnhW@Hf2RsF^Cs$QPR58zNHTC{)kkeY$va??6&=#C)_NpZa
z=Rc71Zc1%F4x#BJF)Zrm5PhzfMj@XU??)~WdH?1le#MWq(KwfS=YGVDP5G_<Pwu)&
zGK|Z9cCAO*SNn%!uMLx%t*)t8Lrx@b#?yZwr}%~R+Dz^vj@7t|fEjeTWk^%Jzhmk3
z+iI93FEqdqjS72f-?99YZB?ongBW~W0;Jx@hRE1UxZ9D_Y~uq1`~(=`KadmAHq}m#
zH|E;4J-IZBxTQE}%=-&U%^EBb4*OL}b9Ws06+C+6y6%T;dE7l-{yd}5EywQW-yyB(
z`bo)XSFlX$>>}}xy5cvbMwy8^Nu_}QY7S^d$lUzgbv^aKTC4BTYCrT+C2Q03sIxb9
z?|^=AwUBrau{wkv1nKOc0e#3@8o~2#jephL+hhi*K3WBY`o<Gl#ofduH29I&E4Qp-
zsb_u&40RZ{T7BYes$G(%_;oym%!uBnW@I-%VU#gXqXa%*;?na6x7X&?Uw1bWo$VRn
z&u@2!M(xzT;V{Hc{zs71K=$_DJ%Cei!KY7O%7aT}<Y$L`RV0TKd#;?UAlU5MMcq2c
z-k6MVm9jalka}^p0h`wOH-p<2*_E_GC-{F;*o)g?1G&RuZC|_Em}{w5(A8G%YximU
zgBMSwe@v^wvgTm2Zl?tl(c)eShXv?3=Q{-akh}*1@XY9(z5$eMZzotx8Mg;L_~MS(
z!)y8<EgPBFVpRI8i1K}#jC`Y}_S3zQ&mKD{u#RCiqYacN+>-J#Y-URU;HSi>YQ5T4
zrC_yG#hdTrX;>#Kp2dT&szo{SHz<FBE60}k(}&buXzkfvLF$5_b41n$vg<rm({mOA
zH5%(r(j-`whthHu++<HgiA_|}^0^5@1v>FWc6fDd-RQ~oFA;fK^qf3jm@xDPpGWX+
zQImW#L<#{NBC-jcjZ3wDkAh4>R&%y=eLTPJRm3fDw=S^?yb`?3%A7`4dI4U*Ja>oO
zG9@E+4soX9a44*Kf38w<UIPcG{J0l#FQ_dt#ua9Vz}`WY9@nvXfol`w02il<BUmY@
zfE5qh?J%Vxtt^`ipda)c%tKnH3J{^wt~Dbi_Iu!))YY(rznpUM5r#jlhO_{kK!G8m
z9UPQNhubyC4+-mh_8`VZax0*AUs)<w4~x4fhA<4#OlFzJddD<3HA_Smn-ZczE7#Gs
z;>4qLR__d>R~AMONnzU$CZHps+8ckawas7LcCXS1IU?@w!ywD0EkhMr&w%hc<v8F^
za{YaEetpu0AD)Bh=Rc)sh#JEu5<tn17DkNk*az}F&9}|3A2?Vgz+EY1f-A<TI=}e`
z(g}Kh@j$U|TtBi}EjG@VXj35bP`^f@Hr2Z<txz+uqNzg9f3e3yTKl{6<J;z{eVp?H
z4YF&8uG=j+I$-1Boyf%pQR#^|uC3z_1j%9e6PRXV`^H@!9(yZ=M((7ZGu&rBu7?Jb
zf^zvvBkHWEbH{6Aum9hL@{4p{nY5kPY6hLWrekU4#8RPCfRg-*&DBvtYqlhoQr>gU
zUptG5vg~tTQ~w}n3&E%s0vOKRh6Ho8`iQ^+S=&Wx^o6==76mpaKC3#Mnb$w?Uf#16
zS9+4DYFP=2VQPl>7B0J3zxzduM3rv>rkQ42y1FBZqH|f})D<NE>E8u969Nq8><yC!
zp^qr|jThh$_U%%@3>lE6!$L4jm%PD1NS-B5FfA8m#S=thJrx@uL3O?!k=qDiM=ZAy
zFGFR${90ax$wLA~{*}Y5fGT9cyPPqYIzQ*-qq`_Ni2-7oLN5PJIUfPkf-_?>x70Ya
zr~h;G_U(;d?h_S6>CN=Ua<026CmFH+*>uHTFwGxbYg~cebw)q}r|6?JQhRVE-<`z9
z7%8z>x`Hx#P6cWElUuTBgb0Ut1~F9886!UtPGoN`?Ka?HMak698@u4%=+{o%>pEC9
zKgg{hd}Gi=H-}${TX=UIPg(;}PL(-MzMQC&e$yIiaK?G#L=DJ6i=0iB&3<6vA!;0g
z!)pEGxvg;nRkbq@N(tM>Yv197Auo3op(6XOQrXBPap_%DfS=?oRyAEwp}knGLzm-J
zw{f%bs{1=J3jqIGSZVc=N~nIhq(@CF+-b(rGe-Jr`(#Xo_1tS#jSf%RY0Ba(augd}
zz_M?W+6SsP004aszEO59V+eLjXQ;GxD&?|Wq&%@rXIY-Lb&Y4~i!%}Kd*j))55yDG
z=zy4RA(|V-)-PdWi!mgPzU3<9hw?##R<*BD?aJ{$D)=yakU#Kvj>{}=$oHq(eni=Z
zu}KWF2s)QDq9Z<xRS0nZhVY&9yqG$RKQe~UalItg?3rcFwhuNncU(GI#_e?DDZb07
zn~uD??x7Daj)rR#y3LQm9AEd!^Qx+*)QvS6yQ_7au#n3)uFJIzTRoErSy8sALKJD$
zMocM&NnDO|2l%`N<dc>;X-s%fx6%?Lz~0I@Cgu5N-R&Adp(6Itl*6aX1U7xho0V_H
z_!FXrteA|qH!MJ@H;?^gm7$ly__@dx!BIGeKwexRpk(&JA=vd}2*UNlzYR}J{=j>R
zc#EB;;hZQO01+AO@Vf~w*PuLaf$-Jz)oV)c$%{0gtlFq?@J17!il-`iK8$|`RP~TJ
z81v1r%nk|{DNZ}KZ$a-bkSnWO0#|&?|KzbBpF`|ybU<bH)cItD*a=m)BlhvFvYU*#
z;0R4$E_L-jl*JM5o(6hP1cZ>Py}?X+T4}z0i^hd(O_GQ)F(@=S1$Z9|!LJ;1P8BM?
zENNS15%tX~>N({GZ^UMkRL*3dR?K0&wIkl33oS8JL!!XO-3lQI9>;)^^I`+9*=e~0
zUgY@tw%G-7ZkT9qu?(|k-G^YnM&{lvd2u4X%5eC6@_WXbde;9f8LPii4FK%^7h~rT
zD+&x?(PzK^Y}>YN+qP}nwr$(C&HrrMoS(@|GK)!Ol5W~nS54F2d(LGL`u8a#ea+K$
z+{%)8QvhhvG<5>}ml$@6)dF4c?50n@Ma9f{6U9yabTQ>&E+)SZvvdS7j0Gy@5lN1E
zesbrFehy#02!nZA21%VCDl_<@y-HuCs9?&lH4sIrngweMo$aPGh3U(W6=)E^?q(|m
zMw+ZNI+V=%od}e!L!gbYshI#V<}Xjpt#MgXV`7tD=BI3#Sob#m%;ii8_U5P)Or*Oh
zl8$>@`P@1iRC|dpUBDAV)|hunk3=0$!Xg7~++tmKl@-aH3f6cVOo5XlX8;gybmibz
z*P-GJYqE^1IFyd6iMTMMC(gr-g#*~fR%4oW0gvD}@%Ag|V$vxF$$eIYk8rKroyCVM
zScdvG50t@}7v3cjdq9a&;C2c6&AKu9kehz^6b!DEQl$#Blb?Gt6D44=hWdsjsW?V6
zDs<Aq@JE_W9fY<-aqXK9MakLZK#;SQ5m+lS$^?}LG8L;Td+;b1GgVF0e6K{(Gf};2
zqWWJ;Z?USJM&teBNYxiT^owP;il1ir1aw6__)*Pu7zjbPWZ`MYf%ZZboSL}4l3oEP
z6Me&zB)R!hJyxnV8An`%R>eJo#hv(YxcCbf-!hPl2w7K*`;vU?e*x*W(coh?Li~wr
z+mme}Pp3G(4`&~aS)|s_N%Er;gK>m8P?G`1+0LKo>C`(z15`HJdJTkS-spIs7;&n5
zl2wmyVg1+q;8Xpr<*meTg-zf~Kv~e+MmD{1-|Ss9m|g6DS(c2NvcL8`u6sLL$jMGd
zWVA&bBXUExSPRrqtF1lLqhb4?)D%7NQ54K1DGLN8sB9}=W>X;<6TQ(CK(FCfEZNYp
z@!m4H@)Dd2u&>&;wfdzQTwZ2^gQPyH4+>$zc;UXLTjMsCu0ickbJU_YL2z6^KsX7a
z?}ef5-&*lF^`k#Ggi-y=*T9JD8;p<FC*Fb*_k)*bw2VC(QH_d))+%nfs#2?57}idE
z66bFc-Xw-2-wqMn1V!PNb(O1o52;2Xe!bu{NXwz_DS0N080!<&!)j9XBf;%|q=c|^
zARbxu0@YVDL)zoq^3aSIXLS+d$a>K>T#t=MApL=26f!Wr<&mPAEfV29whUnBm@WdB
zbLVuVsHnE{o#G-*wv8hD2d6MlE2S%%U8O;3_Smkx<geoH+3#J*I>747(535&%ckas
zo1k^lz<Bum91bj?W>~Wt*>A<G=kVgRs@J{<vM`1B?jGQ$imfbB5E{z5bX4G}3C07h
zOTwj;s-YNOOD+4fPFr6Z?4lBGTKj4E>!#C!B5m+Flt689dN2>5JNZPQ*<N%oQBzw&
z3QRXI{MaamCXr51wi3W|7<2>9e`NA4+}Lrx#<rVP?HOc=^FoszS(^RXq7jvz;2V6r
zrJnue4HvTkZa%Dh>1+Cd7j3pjPf&3)wlUvZ|Awn235AoG@w?|lP*!+;#iu9GYpt54
zu8O<4l+Pm;g;6%=HQRN$<3=6|WfAnQD^y0;#73LMt7&`aePeC&X&k7Du&4HK`r;r^
zc15@hel!&<nf;k(3;st9DK}Pf?iuL#PMTYT_+G6_)NigAnz(Y%+y(VEjNRlVlne4W
z5KTJ+4s-P`#8mEog}X$4>ChM5i!%di^Y&!bpX?lScjV$1Rf^*1$D7Q}NaPelxG_m<
zh;CxK=$NWrFerXScEVwf>H12wybe}8NDO)6zUC&Ahvg|J@k_FTa2RMSDm|*Se%%(D
ziMv-u>2RdM6)q)fuPD+K*9&cN{A!9cNId3(k$GU4ALYK3$DE2#^eCW{V!RC)?hZze
zrrR=|#<+~qB6_%mEE`3GuxA4^3LJQu9N@ntwgG|4ATPCYC-_vV4*op!!a2s^AwgZX
zlb5(I!&J&R7IH0s{EK%Oyf^ZoFsRGR@#6tfXxv#z$Xv&o=RfLz(_+54f6hh4gU+p4
z2ek!j#Dl8mdh3k5fCv>t5ldx5C2v%dLo)%Hmk5(3$^o!`b?}WPc0n-!)`m?prU4ll
z0~A=SODPAdC@<SAKwIPj6rJ!GyT|To<*rd~5QMq=5iHJ25Asi8v5^t@7!GoQBKEJ3
zjbeOGT4Dqst?A_&5wx37AH#6E2aOwzUF(|aYO7}on)%DN3VlR~9*C+CttM^9*YI14
z5f?&^&_;khHZ2dTZa}>WZ(iOt7s-jn-7yu^7(cHCo*VPxnd<tU>re+@rGbX3YpVMF
zl&*jt?$P^pTnSV>Weun!k7po=hi8=2a1Z*qbD+J8Y{F0hZ*b{Ah;|UHS{Ba#K4XNL
zs*`Jd0B3T6*vZl78kcM&o8*hP-RS=t8_@<ksTxHrKI_hv>)&ncvECvSTW5mQA6Srl
zQ32IT+zJ5HE=?C<3{I|YhajNo5KHUo$$RqwyM&P`FeNB|@35f)xCVneA2>soWqXdb
zmfGF2Y4_J+4zVC7Fj2Kunhn;$E2F~r24+LQsrPv;w9<l`gXO6650$+#6G)^&I|LCh
z$nTvOy_Dbqs?xpD2}Ey_69LFIwwWgGp@E}tLs(HLQb3%knh)+tVk?a(ZH#IN@QnOn
zy4QK@0UWic1Nd{`21XxPw%z>sk@^;~KOtR@Pz$d2OSyOOL~$K~7sWrlH<k#>24ap`
zO2|T*kXAPiE$P7N^QXhRh47_2;VyZUFz(0AHq^k|AYJ!1OGITiEpiHe0E&RzhF!le
z%d0nf#}&H=v&$w-dRx0Om_#YN8@r9jbcj9JGk974)}nhw6dB`_$EREsW(}sSVFw*B
zCXUa_jj*imYPe<vy4$mq8WDC#l<IuIC475TPAw;;?c7#Ad4@@I!kTOw)o<y7ArqAa
z=~sE6tl+6}q<R(nhzim9-7Jrd)H${xSznW8b^m%2IFbLC5#33Nsg8oPTgh{6Yd1q5
z5a*~2RLLe1nXay{mn8{L?PbBXK<5y_VESX7YPHpkk-EIn9Zu9GC+Ny8#2IRCY>Kry
zq$@@d*2enPUVd4_RFN02SkR<lf^ql*Y5r7*leP_3SDIG!D*8%ouFp}LhmZ?ATOb?n
zAPM}PjYGvEH#)27`KvnJT|oTfQA)S#?}Wt~3<kB6kP~Kt#hG}KuE&k3)22Zr!dzrW
z=<>YJ_<;h*W<J|-Q>=kFHP&T1@<RbQ7-|<|nQ?LS+dT&Ak`*yC28`eStu;L4u;Eq3
z7`dW`kok)L%csG)z)}Dj!+;H`&Xeg!8mFS%UQ}a&$dmQ_JcH-;MB_OTbx|;;kw7tX
z7H+9_+IiccP_f9=Vm(KXHLz`ukuvW<VZDDH6J6s%7tz)@PKh3u4w9vjPc$Ka-lm}I
zbgf;%Gnk(qWmZmiDtdT5^yrxUrABuwV729usZ^VJqUdeka!b5h!dHuPP`?RNti@??
z8;8aNg2G^S@9N*|)Ch1VHdPbp3)1o@rb2&Ar{ZD<W0NX-KjNO6nS{>V`j0GgP)}4d
z^S#po`P6wEg;o=>xE>!^lc=GtQD+H&@StJ$fN5%KWND^1>F+#un&U9l;l)xNdBS5p
zQmaOajJA}^ot~-1_IMOq_THFzlZ8G->fAAxMZx!mzc$er7^(7J4dIg2zpg7sVlv-8
zYnv<MLG4`e?;?^c9YZ@F*Ec0m1+mwN&3dd*tb5NL*vp~-gp~I>O<GH%{nRMcFFnF~
z*U)Pjz>!RXZv%EbjKf$<d){EDlma)eCODF&01-v;Sv`xkRlQAJh{O+tg82Z}EN-`@
z={}U*$zsRxI_GQ6Mtf*4+zTpDI{&=W-Y5D3fd)~;n@3GMOdB>IQfV!3*0&v5Fse%=
z=>vw*w0ZgN^LFe_a1~v#7_<y(KWal9X;Yef)3^0Ola;&}&LNb(Ehj);^>l<^ePmgi
z@=eGxc-)2O2I@){;=4~xOjCmfqx%<DfP(34QKFR^HTI>Ew^~HFoNtL>ws8TQ^lqBo
z-ldQ}xj4`I$FmE`{ktS{q3se2JVi={CF>TmY1ADGyNgeQHSwmv5JAM))!H){rq#(C
zD-3;o)7%yx;^aO9O0r`7ca>8xy=XqY%kWTn99pb|Np(Nv2a&`Sa*&{|*<766RF(a@
z45G3~47)u&YzNsIj@%;2OWu}<Xk@XD!`#kO)UF7@S<$fZp`_i+zJ5s=?00|8T`F6i
z1w6Gm4Czriq67l@>Xfua12=#z#s?|(XJT4Fq5Z5uDIN$$j<q0{CfC7RQTD=FZXN$#
zFjirKaxc7x;uzuN9J{`&oz0T2gz%hZdd<j@rt}7e^T~>4^-TYM5HY)aYq7l0)pcU6
zVvXM44Tevp_DLmk+65<LRGkq}X<5zDc0adz^`SfKE!6C$SNd<%i2-@T0d<Vfu{I;j
z>#wx4;C_dWLT1W3B*7}Nj9JD`2%h3ZJ7Q-Na%(nbe2NW+J_1?j9j#^Y>`AbQz{F0K
zfcrITg~ZZVinK=!zP<^XOR5p?Ebfla(K6PC23hJz?Q>hBiH#DT)!dw6k2}#cd_AEe
zeJ!7|38)zDQ*k$jrv5p5!<Bw4n<U<Dw-M^&E)|f|Slo-qV3EREGA1cx#Q^b3?7CL7
za%4!)vX094r0nsZ!>&m0LeP$=PSxe{Dhbx{u@2FW72HRb&7t3lQ)4)w4QQx?xaSuv
zmXH)*H6VN~|CjqE=gqwM7eY7^qn{QVhRmPDQiDQ0fNI{~7uS~nx4{>^Rq$;|Ctt=}
zS0Q&o+;GEl_eq}SB!NaIT9C61Zjh6StF%tADawq?J+tR_JLuj?PF*(>dVJ`maTs9n
zORE24H#l*pMimHS+}*^Em<^M!0}L_AS*<S}@JNT=FT{Mrxr}EYIO^GPHQ2f6h$baT
zmQDUdTph{DL=k?M+q*QWBwgQv1P|W6<E@aHPkm4MHOXy|>50JvJAFEIB(3Ha)KFfV
z*Af5Mo{tkOjtzvCd=SH5%!<}N5SD_X4ef|c$}#@B+iALKsn<=<)qQN&j6Z9k-v(-H
zm`X}0aHX$5bj4r9Eqd$nIPLnMG>Lwh3lYi-4*~og-k~2{7+XVExz&y?>jJ<AXwBo$
zQ~Yc&Iol9ZG%N?wzDyl~QmRAD46vZ0KlMNc@d24-L!{tD0SuTC*Lr&s#X0sbBW%Y=
zv=-DNosnh}q_4{;!A<WI_A^8juL9BJ{L9HtY^Ud$GNxXz1e=<qw|96Wa;_h`mQ@%_
z{}jC5pzyVS_^JYLPxBG{Virx~pmIXD)o!pR2Q$NRlJ=qh`(wzR{kT`O2h_3P$+wkD
z%y-l2AUR*3XT{F504<25Sf7NpAAfeQcVBx9F8T=yg#+tzO2AV}#LuVQtV`0}p=1=M
zPS5jgk%EPEkRmS1bRcW|l<$uu|9HMcLD6x}Iu!3T0cYP;tO2g3&_YWL60A)@R_t?Y
zIC17|q>r;&PPQ0^$G~R|plE$TraXs8no@;Tvk>|lEFpQE3%~@?7objBnYW*Xx2$(-
zg6^^+;e7lTQw!$LwW<cKx^KwV++eOOV6`&?lLn0FLHrm7VW)dggnfBo`#yAiweMS~
zLOE-_@%P@SPveI1Sy(a^L94Uot*@XW(m99e$~XXz@cjkxZV?!`0YxEEucDP;AGTyv
zV&nc9orSn-0h488Xw$%zs=HrL_(~^Rkx9nMZ9t1u5--bv=QK139@O<rahO@*WG{2>
zppUuF6BYq~b%g6kiRYjYM?Yn8dFL7}R@TuWmSKT&b{PVPYJ>L>b<bUxGK4DzsVKRj
zFHUR6rTr~!AwKBgbO`UH=OOH@Bbmi8G;M=g499p&ICZqQZhT3JzZdNqTo*@MeS{`q
ziU;tu;0F!mxr$h;alXLw=ke5+tSx&rKFN_0VDMelC_P^0KRG7jamMLf3SZ&mUNzTn
z=H%WurF<sCLR!ae*=?}cfD<<vO#Ru*j=+Ex-+hd9LJe!c-ac?@iU4(aMj}9z<WSc%
zAy44#p@8S6qwSGL;mmN1e0<jHDNy<}5K#j=^b<7wI5~6f&ptEUDHgF$R?X1+%=ZZm
zdkHGwk6pZ-!q+IcnSz)tIokF5A}*;>m3^?T^aG^N-IQL_nCE8ST{Tb#a#n{F1q=BF
z6+mXm_bBd*$t~NId4f9uebVr{2PFEZhC_yn@83AQX!yAt@%GL7i@>E2j~}id5tJ(D
zH}~x+Ll%fYiirV3d^D0lEPtMST*YVmHOnLF)VU&Ex6L`<pK#0v2ID|qj(3!`<HaBo
zRu6B|VS_5%GjeV=bvl1^TdZeAo9y?l8xy=aJA-WhGP^aD5Si#SU&LQZK|5}-n?p<0
z+l^%8IZ^g<U@WoHQuzhc&SSBVoupObyX)G%6bSCD^_Ce@wu@k*m`Lj<M}c=&_M7g|
z4(b?mrhr51+#E34y^f6+9$7iP5ldA2729b|^EuW`u+b_e-To;Oy_s7J{EeO6K&S(D
zaOODckAnW>bI%&4m^Xc~BL-l%8frRzex;I!VCnX<ltSVm+nCNeYt+&c?oYs%hfosQ
z{JX0Qb7JlTl1{4r@sptX56c}Esif)TY?Q~ggP#~Xpv|=(tsJpf7PoSu#u<COnVmUD
zjjUJsUPs8<X{yQV3S;&kO*d0mKXNU#^kj~Q<JZGsHQ%L2OO7ve?v$J?_}ycjFkeXz
zP5WBJvr~G?m0mx1A$27?l`JvC*`~$9)Hkef>W(Xo-m;HiZY7{Joc=a`Q|k%a1~(xa
zaff&qZNmN^ny7V7SH5$0Vc(4Yvx|k$`eb<r;D7J~dBuS{-=~IwnkKV^y-qH+KRSG4
zvv6I|e0gj*q&fXC^5`vE*xiEoHfG77;eMomRT%_Bp$_NZZjm;g20)+VahpjwkA$0C
zPLp~ml4Ey4IoNIV5)=u?9Nkjymgq+)rx&+XP+{qCOOfyhza{53!CSF1<OaPkB0e9H
zw1{!+o0v{Megw2&!mINq`%uLXME;zEEFUOORq;fb2O|(xP;)~_2dyW6c0k|S(ch4a
zm4k_tK){!_f)>J>7zjGD^K?Q+0O%+w<yRn#xic8^duK9`1!aXnA-_1U^uYt~T%8gb
zE`o=CLHu;e&5J%Z%%0BgyMnl{cI)qbLr65t$FR>YX~wn%UwCCeKYgZA!+D{R2hxyO
zgo@7bCV?OYNSHu@>DRrUf?R~S4?a<<KC>f`H?cVk{Tz@g#fv00_v^sZAQRgVsI*jC
z1_TdPv+z7lAXxN~$mxU#Z?x(!@8A>EXa(flXfTx+=e?HY(~tnBIPSSb0LDUQ@<7(W
z25I1t9de9!KU>ga?|5@`(z$8!hAVoG1mMKvt8CwBZyb;0GZA(PO=fLLBV~wSQW_pt
zjp?o)3Q+!#_gn`QBPT^@M-?>4wvdPpm(3CVvu2rhvX(~f(sU{`0ktnzW;MW->406G
zcVMQU?^l<_L1r0ugU<8ynckx|j<M<E$*dAXr{q4BO&UyynK31veH~WZ+wCrjvZe({
zoBx5OV?oCm206Xbi<#o%p4>Y;XUwf6WmE@J7Zjo5>H!cdYtJ)c%fYp#ieTEa{06OH
zrkK;3tyZ938{ACjE4_4y45D)wV1}Y?#a6y)X_Wu#LWzDgN_mrIF{syHlh*RxX<bDn
zbRZ$w5qk@t5*qT|#Kuq)TE4Sv&`?PqZzTq?K{hwB_ahPbI0_OT%a&h2LU^rjK2gNo
zz0D(O<<ZgWXYcw-{g_qB=wj7Q2bv`h4k@N4aHGxl`<Y+w@2*6GcD`erNKIH~g6O<p
z8V=XcMR~pOW|P;FcWO?Ne7AyIw_i<KD9l#db*V|JyG!B!%#w3yK&wL@g?}y4`9>}z
zB-x|M-+Uz1JIZT<tI}?0EwYrB&W8zTo3L;tV~8NO&3ux6ae(;gWc}3krwJ5ZaMQX*
ztwFRgcgl){f}_$p_2YhB=kO14T3<|>#Z0k*h1sABy{lJeJj|=up2pa&!*drlS;&C4
zqZ9@yms%vp5Zbi<=Fy)$Tn3?q{@g=`zYiLcyVQz>1p?ylmGjuTrB~ufBW1RAvVUOf
zl5y)-kj9xD(xB+@p<h>ZwuxHULTJZvQ@O*}$l^+~y8)taohhRQq<);ck*(ZbmELf#
z3ttCcbUS$M16J3027>uS(NJOlqL#TVt;Mt)<GfLXydn~8z3mS4d_+x0v-uH*qMd!X
zA4C%HC8{A^`@L6bVtSN!4TsbEBI<&0<J4x=OpeUf2x>a|O13EWY|ycB!87p}4Nxt}
zz)QS~LiQrVRA?R98ich*87gW8$}<0OnL&J#yT2<vZ2ML)tNGb)qe3&IoqIDmo=tRk
z9kcZ>e)A+4q9?ws)NO{&Y!#PCU<zSDV?QI2%DT4~O`<jiRyz}_D6&=&q6*%xt(48C
zfKYY3YJ0m!!d4B831iIHh1zay@N*#bZk;3vAT4q6I-f5&d4${+J@i(fDM9JfTi?P_
z>!C*?E1-l4NH?0OXr1^qNpV?5MGP7gfo^>$QBdleZGd}RhZ^}beUOP?6em|gwt<gJ
z_DU_6p@5>C?@NL_Pt^B2t{6sKX7@4-f?$4B&)6^b2G`zrO?~b;y(>|bf+8NoImyO*
z5FXz6zIgMHJK#kIhGHNFjkUGgq<g5!)awLChIx6N_V@tegh7g+2qz%b0q#V(KNlo{
z$ELg~x%jsk6d}DtCgUsYNdOdn=@`sIo_G>F&mc%Kc@gEY-9fcmM0LaiNC_$`zh*ta
z4l_tFt5KMrS7BU6Cy}7DApgdiBeA<Qj^G^Zww|Cyjlp0wIWG-7WfeJ_?naVvwZ--p
zM|dm>pmYpEFdgv8MLVm3Ir-IKC=B_V%1fnoaIUppV*cIlCzhX~BYFqse^OuY`%)a4
z?D%Ief`MZTsg_$f`N`dj&l;ZICeeQsd|<MWJ?}pS@oRfkRLeb|@A8R;8W18fFU_Du
z-3vUWZ@z2!_ZUK@(G^xNIN`5=Q+o8`SrrTJ^_G_Ej^qaZC`P_Lb}~j;w{b6FOiXyr
z8GNG~*n;f44|jSLYdi~DwxS#gD_(<Yi$=W!8fa4@BY^igiv7l=l{wB^W<r>~@o3R~
z^G5O9L>#Ox7wtL11Lv`Q=?_OT`M4>RL>o<0)9Hmt4?hrtHYD-n9zhpqO-K6Y3d4tO
z!dMhlOKrz9priEeRQi$A4i-h#<0Q`E@?dR-vcq;|W?d2gd*)Io^$NmbfexwRuW1!)
zr<{7P(cvCE;_3#FKp1V8@4N!1q0%-e_5>p#<BKJI_lE@oYYa;NRkrqb1jF&14TE)1
z<R2@C?sxSzywEj5)@eBMq!7clwR;PrU#Oa)Vq@O2Z5N={N=E#9LAKSWv*9qu0K)aP
zY5M2oQk5_m^rGFQ1-=A1(ydjyCqB<gSaH7Odd`Boy%Wpd0<(V<bxpW!v6sx9A1N3+
zGr<^_QB2JwW<Uig>O4c~BJABgtsKa^oAWcb_Juq4Qi<>)1Z^KJ%w22moO9B<e8H88
z!>ZZ=?@a1z=fuU3Sq1F|SK9jf>wkCROvFHw=R;bDgrmusb|~2_MHc)u_&^>!#uW5$
z3Z}j{DrN0E-+DyzD={^-RjXalP&foJCgYj(vAJB77Dn#30RRp3>9m#kT@VHxP@4Ux
z2u*Jz3z>S2A&b2=YrK}e+>YxIdW<M*x-3)$wCu!ETJPe)J!#IA&~pSvozXGy=dXUy
z&J1Jf%fkz^G+7V^)!>cdSIzieR@RGdkzWy_Id~=d2rM`ZZq%Fno_pc(*BC!E?eHAs
zvf9@P1!!Ig4A^60RQ1A9;$8Zo((NNdI1F7k6*-3VrJgZ|GM?nCviEAtU&e4IMTDfG
zcwvqMHr8CKb6VcfP@(vKODy%@=&(aXbV+fD;%256+qD*Kv7q<Qk?xq&a*&|H81`u`
zUk{6B1cnYTHSXD7<D?IfdR#$3#Hw`Pe?1Va;)};h>URJJJ20o$sXTTvg9pfL5(HX{
zICc;DG5$j?-ilNOA$G-wLYVgp<H5CXX6_I*dc1EA`ecg?sc@yo=Fp#1ABQB7^tl>t
z^RvGwnAsW<+Mvy$F8Wkab*m`wi2y!7YU6ozARP{x^NJtA93*lqk;zVMuEgC0@XsyP
z5wzoX_o_9!2W>A-4~5$GJXju1YdCla8czY2&i3i5-hs%q>t247V{<K@96n+OrCScv
ztca}Y3=5r?roR+oYF{;iH-U)z7v869h^Mxc8ZBlPL%6-&VWimNK2aKPye7R96b{}=
zC&Beqmojl)lfaV-per<6A$z3H%c=Y)?yo+cAajRxNiZ|Hr%~|pu?y`-vxyBXGvb<5
zw!xt)C>OL1miHxB9ZlRURHR%dS_N(E7^TM_c8ck!2!?J54qn$xJc#l;5Alt#vb@e2
zC);$1-<kce;r5;ED+f|8(1%j>?*y*G&RW@~+jDacYh!chCS%~&A|<{Dy?vZYuus7j
zjm3Ff1@n>EXAI=X*hPMmq`oXUXeEuvZ?)JQZXL>`tP2?=+Xu)KdcM}eCe!$B%6S2+
z_mQsUm}8(kdhvBlf2jX)?D>BSa`~r@VUTc}e|lq@^Vpp^!O{ZXP#T&9IT@S4I)@nY
z(_A7>X(~1!l;$CviBaC~m-Nc0+a0S<bL=`8fSpyZL5DY4!wa~l*2)oBr`>Xz6A>6j
z25oXcGvcV3GXb8>UPzE0JKh1k?0VoCgQIP$co}N`FvnSRyw2L9o%!QizX|?1jsq`3
zl>lL~cP}p3UTPi5UwxNat~O`S%C6|%o>7_=<(j3Jc7AEbOd7c*tY4Q6maNk&I)R|M
z*trJvL-zE&3D%!IxuHmc?7ybR{NdYMr)KqW_w<myGq`ihQ|`SF5v#oyHuWZ&cI9oE
z6&;95l>IE+wo@v&FH6dfbbm^-O1`<B(fOHA8sOdUq+%1?1%(i>^y=?^C|+Fyx4qKY
z2C5y91}7>-{fjEjeDK8uLBo0>T^vz^&kQZGC!s>j<J?1#dRydqdV&AI!ryGyK{4Vc
z<;_)z!6Qiu+A`>zvS+TsaJll81D4#(+v#Y_9OcxDgn?9To*IcF4QXD%w9Hbfr<FW)
zBN-Qy{4SrC6|Y+dt4oxLZ#GHoRRZ>vke2;LH|^BbKl5^{vT2^Pi<tT<iW?N!@GJZi
zOyKtz4Xb!|?^kI;rr?H`Z&h05R;ISM%!#$*9jCuy60P4b^%WADv(?%<cC;Wl2s_&q
z1%`uc{tPNAmegjSJHnkmRox`pW$;-XfBtEtBd)f(jZV<b&-;-H5~GNvbZ+)nOoF~H
zhWLi-!Rogr|2po}nRzAyG`aP`L&~8d2fK2WBYx^D*brhPm1>lWlOumNl?yZiYT+b9
zoN&HnCWTLzocW^QX_98*llTTn>GcpSAGRi?`s7$RoC{HmiJ#MLv9AC2qQmd+*Iw=W
zJ??C7jCpyzJGRGM-vPcU<K$O(nH1q-Ks@|o)re^CQvO_@=}fwnU!<!$M%m>!DgO-d
z(1ztMlG#PgJU7B9UQb-H28sMtub;lF>-9k3o4z;l>qzKiI040eyT<)WbNSh1wVUA}
zbSn0)+dzdSR3DPR3>34zn^%X(^I!wx$gu5dnD#AD%8B8~pw@umP>J&tvBu}W((rTW
zce1r+kk*WZPfG1H);yZ`^z}D1a@IYrdM-ALUwixS3w7BH;h1<5Xd>F36(v)NW)tni
zDCZl}BKzIXbS@y>`Jb7ZzeTbw2YqFe{oDH~H42z3jpq&m5_^kE^LTMOF-JJn9x}lz
zCn89d@JZ<97M$6$2XhjH!*SDnaBKRjrOTldP%^c!+@`4=G<my}^jmdE&#M_Y4+1iT
zigGh76%2n#)mQ|yn~+D+gLzB`fh$4$uFx0kP_FPWw_P=yR-a}vasPaN<l)!PQiH`r
zn+@Cw97{^gz3lQoB9y6|=`r@6j0f|;Fak1&?<{0tD$PE3PVzf2+$j-*SJ-a*>x;$R
z!i5>Vbj9Jwa<p?gHEk0+XSyeV7GVRNW5le^rGJhp;T<tLwcyAdU#n0SD~e1N*bR*n
zDq|gaH`u!go1=pLi)?0(I?2-X+~yA;*w|IpSu}~4L7nytHk4i;wn&SV!t_`Qc<FB3
z1q?mbVYM8{9z+(RLjMY8jj&y4=d7skgMBhx*fRry=ca#uYxt;!l^cho>axTjK$|}^
z4uesqeqpyB5<Z~!-Te4Fz&TwYQgFHh_7^2*3J~{ueqU{g*CAFLxytZxqY~eF;So~U
zdVQy8jwct_4i@s4T_Dq{10pS~mHa%B>9)_d%|9kf&>%2_lwWvJOjmd@uvFIdgFDm;
zgcyeXMzSH@N+e1!h3+8ru;K_4z7gluDzF=&9TlL&xxyK2&*GPE(C>5ssU-0{QR<)!
zJF2;H%VMiQmFezus_^4NK))4B9(~>{aQ?K*PO51(@@o{QU!gNE-lGPy0J3h>vVZGV
zzjpP)kMJ&|y|6VeAS^RdxSi=I^p?S0^ohyObxVwXKr{RMOetlJ(+-*NO<|(nkx7H;
z*CBG)2J0^Mqe3=0QP`E8w(Ye78_w+<i5`pizZ!Ta?DCCoJ<xyi=4h_ku7pak6T?=-
zVlvaf#ZBxHO@HCL(BKa<AyKw=oWSgTM~{PAf1WBGgj@)C!GiTR`s(0E!!f_B+qlzW
zobI(30;!yk?zh7j4B7AMZ0<Nzaxz+j02G0T41*^_97Ec~Bytkpev#<Bgz9*1LJyMC
zjubo2(orTr;=9>79z}7Xb(Lx8gqkv-tQ;ywqp;r}WT$TK3vSz{CS+>sCXE4DgEJ9c
z>NWjA-R7_D9J}8Ar4Pz;`0TV8aE`!&fUPW?JD(U2G~{E3m~w|Z7wbn^0QK$NPAhZ}
zHO+79eItk7?!YuH6;4E!n&|c&_RLZZq3gVQ?v04a=*A!V`oX1u4{H`613BnU!yrN=
zKb@L(s;?kzk(GBdW;ZgHuNS3zV79Uyn;Q2TmLfxyZhFSFw%$W?)?dKsO%+J-Hk2;y
zOtZ6ltXcU|sq7u3OGB6_`3~nqy^CmWO+f8Z1PF?jmFBHi%~D0=m#FKi*ZsBNV@ve5
zxt@7Fncd*KyCVuB7iBr}Y(C?zq+XBOGrTBLR7Fzao7FATEm3+DBFv?adO{!_-Sn|>
zPBl=CL@x>Hv|iPea`X!fHeJe35li!<)4}$Rg4i@8mkMo6Mbe#4V>8p<roz4CsO4e_
z4P&a}yG+8!0dj_|Jj?d{s$%Cp>|yQ*PU(Mp&Zk=7zET04t~D740d=#hGvrH+@8%Es
zA`<5^@@zqGBzc4MZPqU=tFb#djsJ@+cq5WdTPn(>42x;*!+G@JQ6N3DeD~@+i6C`Z
zf%w-x-<{65uH$^Le|z@ES)xyw^zZXX%lyPKhJg=3Q5o2<e2>i=-Zoj~y!4DkO}m@`
zaaNooh}zg3Acv5b`I=`J{~YlBwHm_DFy0y!L~98>2aBSwg0{1ATi=kI3wq87^tygo
zHsP^=!~P5w0z#dYQkR6)D1Y7NDLU4_6J``j<x1PG(8a`vU|0F(AouPhX`XWi6g93t
zL2DmW*<pPg!#l`N;MYq!jLfEjDtmt8NI2hOz}QYUHLEu265fh{I55;lZoVmUQ1w&m
z+Brh`*+f`<3{ibByXnsn3N^P%@|wqD%8XIOl5L2eU9kvN_&w`5<SF@@3^oI^=QCQ2
z%eUEEV?D?S(|sC}gK-~f-r4N#2{FR#dJJFPNGcP!AJ$;(8o|U-gAsffK_KEREd$=+
znMAWW$#okA!2*B6;c-H(e0onQmB^O#@U5dH3<p@^%0cvK%H?%_DV&2ilPzHz)76-=
zf{DbmSHR5j^)Sq0Lj$KIy@02pni=;{%GI-&aJ(F^W5t-Fx835T?vT*|kwcwP#|%rE
z8Mqg61VU%iosLoAP)B4PLi^T~jMo-Rar&{i7%Ur^iR<2Ku|QUma7+PCPTd|xosrTt
z@%S6`H?Ey<5qMQ-9Iv16h(Ku{mkW~D;4jZGb%jrD`JZ+WqO|iuW6;9E_Vm@X1r<%S
z#vGq@e4?Uc<xR3VpR%#n*X*bLM-V$Y$!ZeWdUeoDS_*3XsKcbOzib20Kp$EDbkFR!
z82rVcgxoU(J3pGZVftmXz_P$>GBC@~xm;&$Dj0v&K?oN_z=i_*e%5zOFKduFo1S!c
zF9-x=Las-3Qmf6Su#T-GiF&XVG@QSsZPqJ<e5T(!!yL0&EEXFX!aQiHd{`yJcBcC}
zgcN?uj0Ij*e44m=>nR}&C1qDF;0q*PDz6Nzx>#?!#*d(`oQ+3ij+nJh%?rRU;Yi>x
z@MY%lp6#xq$2cxpKY5E3Y9U0!Q7|F@gpmGyB%Z5sj1V>Rg?jhV0V;wL?V8;oxBBvS
zWS2YBtvglBdNhnp8wZj&$dO7VHtWJ3PwFT(FcdK<Bf?s#H7GQ!x0+RDD6_7Y?4ijC
zGNiSHS3+t$$NbrBWYwVWIcMon!{O9)(-dptV*2^BlPlW!pl!<!9J%^xmz9?iz;UZQ
z`GXEd{nxvpPeZt}`aQrea48qZfSCT}%h>+&vMaMNRORQ(g*)@fCYWRNqMU&EnUcVz
zYAFrLqy?WXZtI`m*{73N6-Xe{kqTOw*--@c<FS&``6aOLJk9p=lYX1R8SCbc6`46v
zit==X98-k%LEoG*N0;|Pal<jB-=@7XeSsPHMja7)yyFE~fJ==8lnYc3ZUcwZ&zS>j
z-h1ha#8XIfkCcASWtrd4Lv=41&zFe%ivBD!SzU$y>PbYc(xE1GK9Wwt)UV;-?CLV^
zG_oo^d2AmZ?VM+b`52*Q+G0e-`<P4CuH6HS`(-Mhkk}?A?+9s=xyUlL$l`a{bw>Vv
zTTbKl8CfUT6b^*V?c~o9mTmpzY4^tQ5%ImrN62s)247e@8M%7)ultjF4e1v&iS*<F
zJN^<)h&`FJpcm?!OG79D<umvcrG&k(R_S!RCTgJ88C8p9wgYSovs>~%8cMRq{Hwi6
zLo=}B*6mB;jz3*_*b+iHBT(5FM&&RpNVq=1vmlZx&r92ZRcw)tzO)Z5v<E<1=Yct%
zv^=l3$lnxjdC4WnW2^4Oei{K$N$Rru)IVXuI`nV39Y)n06``gpwKz1=$TKpR_yStY
z$KjL#RHo)l_^?%QM#a!t=A<S@v4lX8y@9+Le`Z5aI*|jw;mjh%{;o`~Nz*7Zk=fVE
ztnf-V*8IEYcZL(hv7DFpb>EMwL1|HuN0v<Qzf5O2WV#9lh^?vj6pAs~(|+CxW%1J*
z6A7k}SfJz>d5BLZak4RHyI(a#X7A)+wXF7p;0C_*Qn~wnUI5h?vf$kI;9w{X=TvH-
z`zr?uS}pzr=+P1UYti&~4{@Uf94&E}ZsmXIkN_5YMYTZ)nnd6!?g}fail6xN4DPCc
zRwdb|u2^++5mrj>f&M_FGLtloKQ!5xM>IS*JIB)j*yrG%Bo8+*YN+M^SZOuJ|8&h7
zSzWswtm<kl^W&2A;Ex?)c$JmIwEf!!h1P(PL2a01qXxj?X7upDkrfzO9N(unQw=qP
zmT!bURCN@v04~x1G#+SzujbOZ%Aze211(h^4H`eFB%q1h95~=m8hZF~7bqx(6uT$e
zi}%m_NHwzYhcaDXg*`46bd~oDzFW&uIRbN6UbY0#SptOLd1Dm$0oR<K&(1F^PYaqe
z3lHt#LH{WPA^If(`1snHD%-oViZ>$y5s{oRlNUYTISw=@aHC1`4Sw;i?+k%4h61xr
zt3FMT20u@twz3V%ZZ`M?bnYM9%c_$emYF0R>K{Z^EoL#M{hVM8keWiPg4Q9I)7*Y^
z>bVG2S^*JmhQQE~e9DO7s@+geCCJCrtIg-o;M?4@PH75NVbUpE&_d51)JY!Ru%Y>w
zd-b42J&dh!pmYb4(1{n)C&GoUbydj$v3R6HE7*{aK0E9Me_v}a*td-LPw=A<ZY@TB
zvC||HO6Qhx_|Z>J4{rg^D@P^#L11QCO-B`L|Bzg<)sK8&+Vo2!@!hrSVf_%3@v$Nr
zI-~EB2?9?~kzzXKW@`!62J%IixAmB77jgX2`!R5QDM%c&4BYL*OMJ*(3b5!C={dxR
zC$H1Y<7gvZ3hHyRS5FKn2^5=5f-u9HMI+ICAnZPW{?Y){q7>Q7d%v@CA5b|bq(lV0
z9@@p@JkGFN2*!_LV5M-^ILBEbayl=U-`i^<EZ66};;O!Ye4Z;xoj1*SlIeZA@lrVL
zmyPIkvW%91bbhs?4q9b7P;S%v=DiIW`5KpN5=Oe~Es)4HV7g1G0`@rFI-TQ4BGpMd
zW|^EcKcnT63K9Q}3!3<0kj=we*BJU0pOANxr9>$l^xWokD;R0^$zn-r{=`GYmF9{V
z7dpeGd7W5xSnUVUB!bUjxPFp0g{_MlE^|4v93eAd*Ld=o464s{n6gAu65<slU@K~>
z5IGg+_Ao}SaFL<E6o6mL)5g|V>DGVlb70ZW)lY!uQMI$$LWVwaNOHulnE3w&63`eS
ze>3EnV^-)6dj%(_j!j%URnD^AtSY^+VPhW{Q$jLlRw}}#sU_r>-)fXmw<MiTjznnX
zm|%bz9^(#o;p(H<GEqI1rb;2Xhy$EGa6o1>B~A5ebGg^qN)AJHU??P7^#~1rPAXWR
zK8NBX?jQ-H{y2YY&?>Qac^^4p6z0G$ZTqyty=L=!LjBf@&b*(*JL)=)F6)(d!GXs#
z2sZ6e7Vq6)(c5MVQTOvH>uM*UI~7w>kgVN2Nl5?%2;gz9DC5y7W?N@>*>ZecY>!dn
zEj5(f3CBjBjW^km_*b&2ZF5itIF7zZDi6=hunr=#inxha{LQA-SigNraEFHnOcSV)
zldzu4c-8K*lMNk;n@zg3<zz<+Xm@|^>ig-rlf=`$WDZR0KH@YJuRpfYU$>xw7F-ck
zhjRFk^xeuFmIJr3{gjbD{L^3;Kxh#dFk7q044xE_ULZoTBVC_22bVd<EBpf*+Q!D?
z@>JvM84q{=Wln;4C^GQt-JRD}EITge&|a9avaN{4CD)-u5-=m|`rI2ct6NGGOz;MN
zi%p9+P9^c|^p(vPzLFb2zHfc`8a~-Z0dV0k&kjn*&`WBGVMpi;*Ve9C+(i*iUiI}<
zZL;fx3p2muUo?G<z~7TcGjMW*y9_qf7Uyiw!1ki4$8YExwoRj-uTYn>V&4>KLQEu^
zGnQi<RqPQYNEP*fU!T8<ALzAQ80&f=U#mcft1%dZ1iepxm7B{l7RW2hi&({DKZLJ~
z90V~7zEzY^%00vqcyE47V5S!2^xJ!i2pn=wCC%AMoC6=#I(D{kbpDvZ>pbBEEGj~T
zWIEo$0u(;rN}u7k5nR%F4bxWN3M^RHsQT{O<5jJ(%c?ZR$?{ct8=Y<PA;^dkEz6|H
zAxXFj0ZuKL!No5Of7n1l%`x6lnakH~eiBl0hR^YQ_m8PV27C-o$_Uz$KUp>J4`-_t
zoDUg@5kztFAP|pOnq-4#mr#Vh?_dK6x$}@O1!^!(h0z@J$2kOt3FHH>E7I+1Z&roa
zHGS>FmtC@9WonWP?_x+>{N>Q`&686KuWTTfqdql8H(HFWDm^7T7s3F5m&aFKXmQtj
z=ECg?+Xgi)c6^U(FO^RQ9e66g|G+4Hu<dE0uclyvgzW@O&cnBFb4m>(XI&sX`CSta
zVi0P3EaBCM%pk1xAUz0m;Zm`W_xN?r!_MyfO7+}{`_PAfVecPS`AkH0T+mfAH^8Wr
zNER>KNqm~%{hY~V*JLcqlS<;u?qpZ*t5`}44JSV}JkH}gggC59@_Y=D++_SQQV5hr
z0u@ZcM|XAHVnk7*Fjbq?2wkAE;Pj(>YO&lziFegBwqs_&Bv+<bU~30%uX!A4!&Od-
zM)*lB_c$C{*jkk5T{?o#JzxA!+z>22l@uU{L#N?g6!JLB0w5aNz`(*70hhN#r&<G*
z&c{hqk)F#P0k|$wT-s5eLGZ8TyRF^%${tNuds;QV(XxJSKQ$Oe>|4U$w1ojVSHi1`
zh+6xwMe=9ev1$9~vndvZwr%?r+?_6#RkW_mQtLKV3B^DWwyT#Y6T@nJN~7$zdp&~r
z&uB)TL{JTllArG~1|$Vl72d`;FfntdknmZNZ-UwhEzX}voYvt?AhdgJXd4)`d2qoR
zlu-6u*-~J(x0r5BVuRD1Tu*xGn&P|<kkizoa+>#pcJEq8$TFBygM8H8*h}C5M`k?$
z5%mNjMsU_5cH>oGgU)hRR8e7F?sXs5dNzVi<7^G_r%bC>9<H}{fHywCMlhN8@`h-G
zo$9XSM&tR<>x;~tYAHdyJ>L+tj`Xt|>)y>Kft2>4NRPZrA7uMw?Qcz-{V97`dn1T?
zUJoArp^P1}b7X0=E<gt2$%L6?0l3@{@sDA!eN@Oaotxru6GSyV0~G54-E=-&(-xJ;
zMW3#2OgM&Uc^$BTfmA+LNR-j`Toc3K*$J<T^<j-H!*nU|t1>RLbh+3Q4<}a)K_bnq
zj4O8&`Pu{f3!+E1O6otK$?zSfdoTrLzk9+@xMTH<&oyc^jSX*V7J(KNv0wbdi*YHC
z=3;2sE^u&lz>xtc4?HVZW_Zc(@T6#JBd44{>xq>z0n0zJm`IB4Klu9`@#NJ{Tzz(Z
zN-GU5pw1H|sBv#&s^<^qc6lLoHcz&2z%tv*gmojrOINIs3kePO3TBq~RKnVyIe(dt
zTg6Db0G1vBQ7NPQQ<pKIa!Kb=^%<_P@n&e6BrT;P9<F^LU;H3%jaBpU!Mg^OyW^JN
zeRd2F(|nH#C`;cs9rMR*a-BA=?c~7@Z%5)x^K@p-KCt?!TcXBBDJ#y&S7Gw;rin8@
zosd)t|Gwtx`&X>I4eHH?X#>g4MWtqYvU~|BfyHdFw=*oia|%DB*N6+fDm2t7{*ExG
zhBk0yUZ=>l1zqkXiXy!)o|@XTh2=%iTBhFwuAcB1XuYTU28hwZyBZ^H>x;>1dS{qV
z2)0D?GZToYCIZzcO!5<5Wk{-$^A<TW=Z&6k?TV5-!TN^Mjm&&B;W|v{*AZQ86WBm6
z!ryb+M!q~l9*O@-J}&UM_h!&+m1Jyj{EhG%vE9RYV4okSk_$E)^6j03K|XTY%ow=r
z1>VEM{;qb>anB@r)rO|}C_e=IJ}qaKdXc~zSS*Uz@k)3s_u-ZppCsdF2VoYU&pdSz
zEW1+wdq!y$986aGvL$Mk^~H7<D^I5Qu$u?mDjRHY<H72Ju^|cxS#WTTw1v_qb_LY@
z<>C5*I2W_2)u|=h{|5Qpw0Sz7gWwsr#_MB;JOVNqLDPzjio{S7<p4B&_;=}gnx_&F
zOELe0^z){7V3D<x@WXo9_)TevK}D#dGFf2p^2|e1&JjTpz8O2(jwkSMicB9xAVa|V
zH<aD3H+LOOR`-G!Am_Abk#e-n5Jz-)|E8wsE-=OEw>weM%;UP+p|bgu{%dBlOo4K9
zF3BNQ2(Okondc&iW3pq#<Dfu?k9OuL^B@}ulin$Oouvd2T*@-v*x<TTxYgL#jr`V@
z(}_Sou0my>o4QvU-)?+23qb488xIkH{}!y~z|Wu9(J9>JZ0a<x$`t`sp8@V5n}Z-*
zN&M@ombr5|g$@#$#;te)ICK0R?p(SP4>s`q-_<7a>=wVj9jAuzUroPar0+x~V`~$>
zMO-(BFYz1XcOn)|1sIU^jOg4S(=^54TD9S^t*9_akw?XCA}&#Hk_TpMKtc)_8lpR2
zlUrxCiE7Ze#*wO6R3qnX#?WIAmo!D!Ho;#=&#da|Sd6U|S0P>T;$y%o(6Y5v`Ze`L
zE|RC@`@xOT{`3R$J{3OZl9^wsU{o;=?Ut=9%)HUJMUhmMdkKB6jVN3Ni6)TO41-IS
z)kJ=c2>|v7S}Kww2xdt;f2Te*QBONLf?bixWlTj=;+EUKt{@rAUJ~;~&c0Q0NFNNJ
zMKtN+7Xe-q{I2GW;;!#eTYYV7z-MDnhO;j=Q)lu<pYvM}g*?FG#i^)f^0&t3;)jk*
z`R;y$s(iy1YoeU>dXvW3%kG|9wSs*G`OOYXs&WZ`Gkd7E=`a)L2UU;=L&|v2RcqB(
z5N}{A?pJESr?Hh9=36JKaXq3$<h+UzA&*hX0wB0PK6bLbQ95w6S~(RA>vfoNIQ@_r
zLb62+JcqKwLvsF@wqAUGMQQ7?JzZfR)FI89L_*1#=p^hEZgjgUTciH6Ca#(z+$<w7
z1;#B+a_S*?=yONER08g*-dv~gY2<<l$17|J(7&q!QM`Bak+CH>n16l{LP19mUDqkO
zu`_2ffvqFnH*Sl00J0PsEW8?zxab`8tsY=0T=n#*d$uO>+iCj1b?UWCAoz!Nk>fSU
zrb$B4HS<e}VC(5cRj7X8G(XLfr@Q_3{o|Yry<%;-GaV3oEc5iQLb#nv5#s)V_CmsY
zZeDNht?XObKp(gA_YXO<hJOrE6TpNkIsO9Hk$5#Ko*|&*arA})2Z9jL(D(*pbuA)G
zFR>g@gG;(8E>zfH;h0D-u{l0taVfrsH6=hA@&QzrS{wpu8Ha|;BmwU(JM*mc3k$OP
zU+_L0{}<kek(r&9{lA$${}b=S$imFX{Qt%Kv>PRp>on46ZkqgUwfV16`=67nTdwW4
z&seWftamb1{IzI|gp7v7VE*OVRgI0Io#24bfYe+hJqcxn(16@r+`QaW$i!gb@rB|3
zjlo&f_5O`uB)yqobiIjT+`IsUyoB6bz(f$e@!6fBg~919jJ&Pz13gJ^6)RvPXe?+k
zZxt>pE;VQ(CvRIdH8o!~GBiywG~G?Vz+`FZyPO%BI~kg4nY)mqBNk%gk-LczW0kuR
zX=@p5*y{UZ9he%L=(`e{Xld&lY4SUnYwY`)-=2Q9#wM25H#}+C=3gNFOUr|^FF1N7
zF)l1`3<MKE1r7}pZv+BP69R8wEhBG>d%u65*E#IJ_pA(#&F^_v<@Y(iR2o=cYK&iX
zPkL2FpM4~n;$UFjM$klN-a<%B&|2c1zxy(StIL1#o_8&c&^~|De*ll3;CQ_s{Md{8
z*;pCA$Nr9NYOeN=g5+tB&yMblYz!mljn1wQFQDoz%#QZI@aclFqORip_WAyf2@P)S
z;o*W<&-@s~{_IjhlPN&EAsKl<pZ=r!u+;zjqx+n%NdM@dVIUbnqh<MxWql2e@u9G^
zHnqIG@oD`ejrOq?D=DTRrXZGn*a3e{Q5#wrn^_xJh1MN?!KH!4j{d6rTeQ=^zrA^f
zOaFC7;vM?cnL=mh(riDpT3Gs#^P2yid&ZME{q=-|+U)RJps#n@TcT%R++(6|sy}d|
zZ+QRWvyain#*wkk`4xHN>-aMJt`lWsaA0Ey)YfZYI)G}Gwt}s>L!rp5`VwMcE2<N0
z83btO+;*x8(PuEWt#)&cdC!l|xCM)0!qOjaa%-WwpBFTTHv{rv<Uo;uuWbr^R~1yY
zIMM3*E|9<5fv@!fE{rrVB8GOva=6WIXuu_x;~2?K@lv2VmmdwnABC^c+gbPp_a*bS
zxt~yt77T-7P<jO4Jo2vu^PS}`z8j6dweiCtr{*(<`$z>5n(MCrmlY<LJ{-6TgOs#c
zn{QpS7@$#FluUj&LO*1u4lH8-+6U+Dm<}|~bT^WSadJp~4`1f}m|bmqtvb~Z!oc21
z#Xo@b*>S@7(VS9bU<E2qRD!$7c(5;nioTznJj1V~i8qvcI%i89sq`YVPeUAvOLYd)
z*AR*>#xLhiQrT*A?8^+`WbnE~kG>ny9LhxgC;BoezF3j-0#gNWRiUT;1%$_-C0*&L
z*3=2>nsTU&dGqlr#p-rVYC@OqI+@sSvF!P&ovp~0ww1_-Ci)KmH$ce0yc8`X#8%86
zMM+>q!Fpn#cB6V*?!{j`|0KTA9#u`Q70P*X_x$NFhhg$DA40KA!RaaAYY7bB7|8%|
zM}3JhdDM^)vI**;sj|@O(gy|Q5GmP+y0sBkh3G>Uwx@bq0}8U=qDNP0R|;UY@6EI}
z&yb5DhLDcYeJlfoD?I+A#cqNhVzVhC3d<&eR*NH(r{j}<e#<V1A{6&+B#QMnJkI~(
zJk+89JMMTwZQ&l(GAo)VCx-=G3%h?Q;<6@;63pS9wzj2r@M=O%na&i1I{|Lbk$wT_
zT-G_-cCO)MyeoVS1>H2q`(-8iGkZ#T$Oe(hW`yRF55H%M(u^m8+A}J^&O?{28sxVl
zT7k@?wkTMu;WEGm1doLD3<5BUeT)L=G7JyrYqns_MkR)I+E9@?+dUBFD?0)|YZQY<
z0q{e>92_U9uE?Dn`ecCtC>fw<@p6STfDvh-vKBT*`_!RAB~^r67>_a7Gx|@C$Fbwq
zVQ+wWFB$)Lo=+vpj%gofxd`r4#LepBPYoQBu|70o@=rJ-+5JPldUjiYR!7)No|)Ie
zRC3uHkm%D?m}?rozF*kOWLFtPb%ebDesc<{E}ba-VNVo^Nd*^KQR---+wlQUOY5@`
zK<laXWHg24M4pMV=f%0~Yg19<LR6}eE88`9-l#&9qM*?{Y6yZKp)@>4VPqRO7;`Ja
zO;3teIUdp8``hF1v4>MW1cXl85l7jo%QPy5xl1B(d~uI@7DGD4iM@t&eSFLAAc<EV
zf_JE^fpkf@Y%S!6<4(8h2a?UYJ^V{-)$0kZBTXIW(c+&-Lq7&|zDoIHSwr?TlS){b
z;clXp=N?w^JV~C_0buqN7T;Cwm$nY$yGj{VJ8dO0-lRRNiCl;w$E_e(dNkA!>Q*Qj
zJh)kj$(-hR%jdX(!n?UYANh95q^+))WeKk4+~csBjY`-=>BHVnAM#nO#pazn<;Myl
z^t!WcoCkF#=$^Q&w5+ooZMR;;9DR0MlD%7IcwMYC+RpiThuu|HoAjoSd`z)$xAA@(
zU0LqX?nxV5f^W?rK(WAVem=P23iJ(eHtF$ZzB6H}FL_Z$h~-$=sIbUG2`O2j)F+Yr
zwKX3~;aheQm#*K;CL%T2<fkYS0+;TFe$@iDJ7J_|+*nk^<=6d^47aINEgI1ntUH_w
z@0M^MO8#CPMszU3ppk~AL(8J=#-LsYe*oy^5IOOA)$#2kbS`oll7V5@a&v470H)UD
zC?)KoW%4nQJ^f<#leU%AWkWkFKs*Aa%7#OMpl2}?2JEr6>Gf3H$3^1Di1}149P1>x
zBn}<aDNV8N0YHZCr80;GNW6eqDUX`C8V|mTBP-9U&9gBuqAv{Auwz9!&CJxACl)Q$
zsT1R<_}AW-C$^g2X1j4c7N6ya!^>mueJo*XeaAdQKfm^3`}yRJWhX`Z6SOQ{nxOAr
zeWcPt##~<?Qa{cSepOw0>nH>4K*Ubox<qR-(a=bkHO0?8Ry?uRKZjH|BZ|<=cznb%
zN6Z-HpT#3RX<*T)BZu(2VG8-bn;SB6ZYZNpI#`H*3Dk*S;wX?0b4SZrvMTjgaMGXg
zdL^Ly!1G=$DB_DTz1>S8SzTwegF|C~xgsi`2ASJK!omQ>L=ykPt4a(gB?+)=AUSqy
z{u0HX6s8H#G`paFppfF%8*SD<2Yi%=l!xz&v!paez382wi_cs3%R;P3><{U4KBJ!C
z{q2FS>UfL0q&PS4ChbSU7-TrK4#Ix?*wJtXCyx_A1~`sWn%8n4<{}OC;lhosTJ%-L
znfAVU>l6!-W$6n@27#)27h%t`{@@;r#_l@Muv!D_ti}QK@X)7QcZE(=nO)v#7c)Qe
zsynQS3g9V3%VKOcBY26hNtOBT{^^DC2>vsA!(2VT9+4Ps(a7G>Mq`(^s*MAT7(NK<
z6g?|o+*BE1%NDhn`T_jIF;2rq9;lQ}0jeV`S-o;|BPsW!#1B$|(hpvgF)EDi5+%eE
zfJzdQ)W|8`Ubi<w^}#6wmgRs4hms_^LG?G57ZcUcwg(YWZg;_KVJmOW4HzRm#0pt$
z|CkPwyt?ekv1>9t*-L-Rfbu&c3t)Ca$CO7ZNgQ+pS@0W6Rjmcf$m*Db-O2iq#G+pR
zv$Jct06=*OLasoAN#}!A_KM|4xEN^90w~-|jB%jZT<*1bgq__hT1|1iY%THhdKme@
zd@Fj@Sc+SpuFvjadknPFy7_tp@4eF6p5XJwX5*t;<#e;2iyywKVb{=w?G<q%WqQ~`
zyb(n=x1-@I&f&L$8Lt`E{?);cEcZg+@Pl&<EvR`zAs^tx>>FTDPVmqE<~l%`bPv32
z8nr^ee8l%DS8_c|64|uFf_UNx5T-lSO0_})$31jW4-kyNY=Qdd&GD4bIqH$-S7a_&
zYf9rutBd^H5&CiN81=7QqI3y^XQuHXGO>|h_)oTu?x}k%T?xQEyJwBlX&)g>Fn=g@
zTX>i4`92zLaEg`&w2RFyG#>gch-%m{YEvdggikFDfK2(~@g(;fszavIWus4bn)sg_
zEpF5D0GVKlr;`aTDx0dQ4AobiePZBZAsuwdPH~Op$%*NFiP5=taAYmYH-YtP{ytIJ
z6emKdOYtfeNSo&)5RBX{+*<sz9-T1+cXwlZj{ENd?ruovRBk~@C*zdw4JWU2gq2Ri
zjh;rxuIIeP2(_>olfwK;nPd@@uY0+~?d=CIIOw9hHA8!X_?s_D%8~itbl+h3M>cNP
zM4I)g9O4O3=v?nT>7Fb7P-KnBa-Uu@J-cqM#|P142q`NmUgtj>M)i6kn1B^Yt;X2L
zj03GSV><=|ftXGka5an9+HHy5@-4>2?(UP~jN6puohR*R&9*ulwAGNuot&40;?3Eu
zmw5|vV2#iGe?c#Wk$`M=mjnk9Yg7R<ERySpY2^a6RM=!hO18R>8OKB1N&16Uli&r2
zE{*6D87vStTVutl<rPv-IeJfOZ2?3q#Ji*g4QgCbi+p1aUf+bm-aJ9GxS*77KO<k`
zh2tRu7?QVfH8$|eZ;vPj9AFX?dkax;Z9~eZQaR5f+anIm{WhdJ+>zd6y4OY6x6GOr
z8A#c~@+9Fe-!g<x#ClMCzc$>fe2n`gAmtV1-#{U>gKr#71~g<8Y7&GWiva*DxXe>g
zj_?E;g%L?99!IF)T=-!Sl%~ObgOO<6qg5|qX2kSO5o5FuVg6*GD=uyLQ(w*;%1=OV
zn6`Xe00(n5fA?}(_UB_eWBTn5`DoxOIw$zG@QJ%BqbQ78+*(87Zhlo8zg{UWX7JdQ
zTg5DZ^#cJJdDGhv_qqIK?vQ~;A>y?0_{H7%;pGt|Ndo$`;=2T@<~*MIUA;tzv>G!{
zKKYSm33$D)oeC(Mqk1=_fmK`!%+h}Lihb^v>aBqtHhJCgOUKkhiIUqM%uq#`el)`M
z{b<HoU<!ph+%bSlAYeFCNvy^Uv59vR1wza>9Aq6yD#%NSh{Q~EXsWIfQ__u*y26U&
zgCpd%gzk$>w5V0n2Y@RFS%O(|a%y}?*K0dCgN0OL{tEAGS@uH94hK-QLTpmWp~V6w
zg3d@qdE|O`KvzxabM2BpOftXY@z2!H^T{0lIwXfN*c~QiQyqql#UspA(V%009@AcR
zF7``<(uN}Lm6bs#DjswErnK3q(7auXxzs~<QvWi=GC8ajBZ3E#M8jOVHJ{SrriM@r
zZ?(DOgK_o%a<@{*;S|>)@L%iC;F)>=($N!8R~jnr^iI+icIvt2oUBtXAuy6;IKdxF
z_k13S!X4CwE6R#~1Y#K3T|2+(`fG(y%>8Y=`~5(a@Avr^o;Hw9rmDBSd!jhBfe#n0
zQ;d8JEzQD6H(z`*yWB8%xdFucR#>j=tphj^cK-`F(p&A<8r6{<X-94wy_mQ&CAANc
zcx8qn31*6*3`CvpClSHKM`8{*XYrnoIv9GIY)9l|#2b9jP=#S0U)a!`Hez+Wd^<jx
zu_W?=Kxisq`dolP(P18rJedzm%haZm))huuT8bbp%Uf1D$X4h5Q~}C8c950dA3+VH
z>B*ijPrS5qYm7vSS6?S?Y{vn}-Ps7=ND3v1;s>WeOw;_R=ej#&TUVHkl&EKYEX)vg
z2IIvyAr*{P_Tc#OZY5f@eEb>A3mC2vRIw1WtZImJd!`WivP?U@3YG*IHNi@i&&G2S
zsHA+-WppB-K83ZGFh2>%!2>v{cRIG+-9}VPQbLR~dX3(!w3#d4JNe1^(JAA#8Yp+_
z0lmT+)6P$6v+~P6kx)P(>tIh^>axVNMEw5w#=G8MWv?E%O<>~GWhojJ*a`Ae(?ENj
zoKnLL9y}>KM;_kBTW|GyGeBg94O?`PPj!s7%_y@%r@x<8by-^qiL1uOE;7@85@N#<
zRJ}&4J@>eIY{~hhF_{#v&OSgO>C!{biwBgiScLFFEd%bXZdC$s4LD|!Mp(ai4+Cnz
z(?jh>y%RJIbobzv$A=e>z>aB-LS0O<Les7~wAXVk0U(x%xte`V5ik7Ytg^F1fR`q3
zQU=%4o`a9WoXK{abL@F0Z4q`+#QCPM!JQ3J_m2W1*te~pwoXLYB6Sn2R{(LW7ac#5
zwADZiU<H;1@I^-_yozkvlkqi;Hjw~EY0e0?lvu#X=*@Uaa#{K4FC`mSH^iWvQA{9#
z8x?yJ2N~!f^1<YlMS2Wl9|q#6ack(?>}OO>t4xh3P9u23L7}@FMsCT|rUOoLPdicc
zC_B0)VMZ9BqNyR+YzJ>5f>1b|3}CPH$@4z45B{XqoqlLg3x!G^+WObIhmkZdTj#K6
z(@5ReGcGO}Aoj@j4~b3uKISDxKcvMU6Ox=eD-)5pyL48?xzlkUl`&5@bIG}iiDX_y
zB0(3{WT4&V+6R6}XN3FpUe}C<1d-~3$}<TSd_E@)H7Tl{1*K?cJkZQtvCs7A1Nx{5
z%HL5s)1*zifN88QNg}!3ABnt|`#|<s?@4I1m3rx*&)A3Xh5j2ZQiiu<ba?~xH4YhW
z*%VeG?PpJJqqZ{DRJITp$|Jb#VJxC*z(+6pQtWr0xz(MjGelr33FW>q<P&_Hs{mGd
zD&tu|q|oB=8({)QMmaS2MzEl9^47if%j}>ZX4sxM;G&3lE9QcN!LbK`Cm88O7?gd_
z7$$Yd0?jdxfJ`EJw_9fzhB&nUebXVsP%}NHuiM<mu)1hEWhYP3O5yBrd~O+X|B9Z%
zZsYT8x0tDa#sR>zb}C!}o}DqZr=J;1u&O{pZQq!pQp-O8GF00A8?apVkNZPCJh9|o
z8W{Op)JafbA~@v*yR01ewcbAWJ@_Sq?<1CL9bd(v#=`OlI<EwJjTTzyOVwRL!juIW
z3<dC50vYWo??ive)HJ%;F8h{$Do{2B8#JQD>q{Ke@&_;{mts1xLy2~b=d<XUjQR{3
zhE--wd2ZWdHTw`}2VkzVOG^Q%*;p_^>I>gB#m;Te=dYkzix*E;B742Lj`e(_)d|zv
zVVSrKBphW4%~%!fRQ+E{eE@Nc)=l~P69ax<QjLCh3SEb_H@qGfc;&Le;8=sFXt`-D
z(ztUD=BlK~EkDclV7+=g_oShVG>5BA<|rkPrR~nmmmIq&iEhHVw#Qw1LF$TO3MDBh
zd`l1WdeH{XRZA>K$VJ!Z3~JRDAmXzdjZ(h6*Ol4h*Ika9YPmP}DK^GL;@VuU5z{xb
zWR~waDEy+^!>UDx{SwR-#9Iomz!uG{G+QtnWCo(qWH|4}`Ei&dq?XR(zjBDTWsLwH
zKJg#MQS%KxEENfE?hCc4UliRmyMg-ZSFwZkWj?`vPE>~<Z2@)S=P)ygXV&9GSw6)<
zp%rQYpu+?`ylJ2tYZz9{?&eR1QN9$c<HxQj@9L8pq}NvfahN230^;*3^^_3aO(Bv>
z;ssaQPp0Sf%B;epLvXLp0o*&%0jzV12zZN4Smw5#`D7$JA`ZarvI)Mn{tyP%qWLDN
zEh-V13hGSc-?_*zp>j&I_U5o&R=t){;Lx(I@TMKWp4#jI^nzD`1K;i$wUy7hriFz<
zh3!RdOIESlrd*=NZpQUeUX7HCcql5Nq^onsqVXK){%I}x#u>@Y8_oJ>{;RDDfHwk-
zw7f0eG7pjXUJbv!@V(B`^rG6jgq8e!z)z#kA<cvo1li>DNG&fB4ef&ai^i}nBA|1<
zpaC|iS5^01e~8XOiZ{W%0$&ds7@Oau_zJ@;1SI}f?l-FSwSk4Xevhlk*}yBze)KI~
z{ZU!sz|9((4p2_mL>F%i$0}>r`PP;ZvQaz6HG6e8))~4j6({t!^xioNqf>xE`D^wp
zJWNlPD>E)#x<EL4xH#9<Op08m7WOemvDSibrn2T6iU0)G|IJfr?HxA%8R~xxVX&WC
z2Wo!j;o)9JadV<O_E-I3fQ(~b7~}3#uwNI3>tD6K@fzQhJ{qjj79kR{>hnq>Q!~65
zs#TS#e<S`H91>{Hw~ROp(1HJ2gD;J0;)_u<uVa%<Wo{9SCVODeCd{qB$WT%Oi~VUl
zr9e`9P;#I4Gf{`r_~u=_Cn)&>_DFMY@C;qS_yNfzAI_gHe;WNclNNMvT?Kj?AJPf5
z2=I-k^BdEo0lw=5>gCAUTkTVp%=Jed*M7;Gl}>Z#1-YgN*Z<omAZG~a-DB3R!nKZC
zI7v7Z5b7~=@;;nrZs4b>q4iWnn&lFC!RqPr%gA=se{Lbi&{s>r1z_G~e8aMO09VmR
z1{gk(ej>=O7LAeCJg$dTcIa=3!fYSS7RN7OjN<Qs5z`pbWUTUnWl){&Y~pMcl%aoe
zBc$CA9>xOG*~lW1MG<m!`xF)!m7$If!sUQAc)>HiV<MpFnWgx4P4rI{jHv3WtOB43
zzqV90x>|8-oIL;E6IYe+*JO;z=;F}TSVOI+HyawAjEpfM$u9XAv3kf)JGin`f=hD+
zry7F{L=%^Xg7p)LHRiocZ&&_$y5-3=UM~77!7PqZef!~`xU}QuGdRmCLAOB~|I)5-
zKOU+=unuyww+6L??@^{)F>*5%gu;h?C3ogOY3IOA4FuqV$iyEekS=Bw`5~3{gtsyX
zKoi)h7tyLu#huEK&pADS<Rjg}nqq=22xF87NT1+4GJc!Yd<5fR!;cYBCISY7gO^Pv
zy2t8a*F(f#o7Nq<0A*%>&d`7q;LZkau^!2SVd-%TUvLHmzlFXOly?s0&k-8{s_>Av
zCeX*9k7QMMYxE`A2yd709=|ktqkWu;lQwjdgp6Q6CW}rfGbJtUd!3VCFL2MYa-WKU
z-8Kf@pZnlSQQoTkw!(Q)3KXeNMb^fRyUa+Okn+%Y$w0qE30M$cssH)~Ly!1^Rp1mg
z93MAbV~~&tJQePOY(66tTLG8+<xcL9bax;V4_jCH>#YFCNqNn63K_vp#+}KS-@^KJ
z4C$-eE<ewinPp!fonFkAsyD7uy{pW*RsrEd3fX?r2SUyVqXytN4=?$Z<^nlZ@oovP
zmCsXCi}pNK*6Apr4Uyj26%;<>2Q~9;NOTF2h=<jY96XM}S3l&$=n=Xv^p@O+tBx-p
zvXv}?er{`RSH4OB&IrC7M<mup_-7V*<x&M%if7p9@Q?NJ4=Mb?)iU_CJDMF*K;&}i
zt?uXai0#J4pfUD@6m5ys`_Nq*N5x@-BJuB-6_52{4L@ZsA*kXub%nGFFM-t{UuV<s
zBY2yMFW`tAfV@54-Ry04{IhFUp;>CDQ6o(y>)r;U&*{S^=i*h1mWqOWO()Bv1xbxl
z1DuFIfDe{sjwZApDdG<eOOeEk{`>Hfuqs>C0a#qj>-vXJR+KQ7ppR9DfOti9ybs|l
zxYrE1<7~L@!6n!T^J<`w)ZzKM&GIQ$mZrWDCb_jJ{oQu7%XTpj{UA5jJ0ZvkQ9j*=
zJVC1$9M5w8qM9gxoTpX%Y|1sHj5@F$lh)?tnMm{(R)ZC$G0!F3)Guo&9FeVmW>3;H
zvq?~78dL^Dgb&HS{xT?C;AxxcX7LmTlp}>+<f3!j=h4v8g~x8;^gxApN&J>CwykI4
z^<AK+o|VfHwqNs_0cx0l6*Y~T>d(_5Taz%yv<JI`X<Fb$n~h(xqiTjii;$1H?~ddK
zL@!_*OMKWYvTh{#JE&}+LJafdX>hNPU4;gq9BigV4#Iw~ee~`^O$aU6Q3;icMi9cE
zo;j#Bh0+GW#HwW<In#7AQEVJ{gUjR|g(i{r8~FMk56R7<NnJr<q{$G8>}@Z2XN*D4
zsrKOCY<B-Azstxitb@A|5E`aTy-zt_Qt}W+x8;LOHV@I7_?W3Sa$mI0;<uMgUe`#k
zJ<pio$or=^Ri&B)<@Wm;^D3Wy&>2`G9Msh&^)+I_z;-EZwdOaC-_H`$Z(mUjPZ%OC
zYi0C74^sGIS;T)HJ@qQ|k<2ekHHpeK`Ndr4QlTIxg`yuhD9J#kQCO+5f4;6!>8fu6
zIe$C__?%->hr@!{8hM_Ffjj$doS~PtCDWpI9W+73HWf2yPSAEF%vt^X_C5vLlWUOR
zDTYe0%*`&H?9iYF56_Ma?K<18*qKv4JFu-aH?lP=_KkW4m`!^k4LWIK=|Ts)!ACHg
zj#CfdHv24_f+30Vo;z}6k*1Ra2{-}laTUW7jZq;o1Y)e*WFZcLkK&Jj#ubo|qGXSw
zS@$qY<r0@oXzKj)#m(iB6fr^fZ&bNRd<<O!fZO{hI$|Wfk}oUe+RmYbhU27PZ1Dxn
zyX5i)^9Rv9H%PT!p=PxCMZJbligo=NXiriIVC7MECcCje?z>h`Aut-UyARV$W+T;i
zq8avB;q0{4*YB1nz}*xQF&eQRswSz;4dhqv4uQ`jaOf1MHl={qv8X!jqTN9R-H&ru
z<dad93q7JQDpczoJgL@$*<}q>ArFSpyi05zz^u^_x~I(4Qq9?M&zZc@*V18M=avff
z4^sq}FtK2gpv*~d)sKl{P<~ATNb!#mxm#Zvn07lfc}-|drX4fNFnJuhm%YfcZL`Z`
zT>`;8_HDxpjSRD`*)%&i>iG}h^GTS6{lx<dXHlpuBzRXZUwXX+5_03RfJL&VN1?*t
zv%=)&(eK1LM>~^{ljnR@l64&sQT1(%b*%8VeENI`=IF4M7tmI^L2$!I)k&>;E}weH
zj*eSsVLYh%%qx`fVM4RjQGE_L=yrf<Gn?uDHrZsD2ZU<+-#XnhNcBGoPu{Zc(*di;
z`@F9=rrfD5Z%_kikLD|mdI+?MKobWqybC0ll-T50#1Q1;zmP~9dSVT<$r5tC|9PdY
zlLg^4rZFwm?ZUQJu}3IoO7L^|k)1YDrEE)h8YwUF?(F!BV~8QL+mc<Feu-39c~S)v
zcQFve%dL!b!s4~_*akq&fTlTvVm#_Eu`M+%$pgZrOr)l?xU^Ur<r$3P10T683OM|J
zkBL}Yy|LSck2=Vk)WO*iGsu6}@SSAYvL2AE;!lC=NR#0irt+61oWiram7>(uNCnfj
z@$X_&IPdI?L4!9Bno4Dl0~s^DEC29untT{JC9mq3P{z;iM1OOFW=A42b{~Ov4xYoP
zU5}hxs?;U>@;wz!=)|GJ`7<23ndAUtRG~vDPu;}FJ1{?%4Hh(ya}1a?<nER!JACrs
z6njP%Ec3{MzNM5Kmu&Uy8A=>1WCk>?`}%2B73)tHOfbRjeob?3XMVY}<*nFRa<`m<
zL^Hz4+!DfuA7R3VP9-^$^Ta#jl1;dds4_87E_9;E%<vd+nkC-00&MADNON5Zey}_7
z4tc>IWsOEFdGe_dsUj^$I@)svq|e2}KF(r`eu`m(J|6T$k?Kd-R$-3^OF{mGjNRv3
zpBSHvXx_C@LNEbx6=Q-gs2fsBN20V)p9k}CW`Y62QFc+%Vue$Jr>o33&GibUesdlQ
z%S2%^d<ve?w1Mh7qXzLnA#)X4uDb?KJF?FY$1|Y0rK}li*-_LiB<1GNsJ?q3$Xqyf
zij94^PaF*$j`b;w*?WoE-5p3ucG2?Wp2nh~er<T&^&qmqOoM4#c-$NiYz05@{Z}JX
z#X8}FkL=@Q5b5(rLP0?kp)EJ%cjh0`@C<)?EXKYI$nk{SN7C#OA#cR-h>v|Ep&6<r
zhssk+XS}?u^`&Z+-WC?ex<jW|Ao=|~fHc@&ArYdNXh2dctFA*LkF`8<fTLensBO~C
z$$Puq<U!5f40ppT%(WLQi=VXRDLo^#I+M}Og_uot+k(I;M1bVpn_3IaY7rS664<hL
zC)Z9r{F`eRv-G;IWnn@DO&3e@QT6U5H@0iLk{FOw1syCJEO|y|s^X_%uicYwFg_yJ
z!`z0C#$ZUdQc|k(&T&{?A0RM{Ipg6bTQIjYBGuMFBw^OS`79i3oB%++;c8V;#|yV8
zcqnpqb6nKU?P!lMd_s$=D@Sj*epGR>aIDHQ$#PE~TlJjiKT^v(Z9&%4fP|FVmvUDw
zz3w#nSfqScix*g$*0!%N42D7DKcc{8vl46b2*VRRt5RSVfJ;eJ>T4V6G8bjiFfT8N
zDKRCao0NO<=v)@Bo@{!m>Va$62jvV7vs4S&L|`I9NLp3ROi#DpRH+rTdQA1oYVyy|
zFAtDtN$1_N`cW@WNz?bwQClH=(GX-@Qr`fSQiG&T^L}+ih+&zT>jV;+VSP0SJe+(*
zLRR!V(Cji!lL4EBmJSAA$=rg!Q)nB3oUN%5+YVuO#@OXxjT+Hht7V|C5AYI6@t&p_
z^OLuRZ!}Qye^*mkmL;W+db3F_s3`$2Y{0s*diA7XOU;B#sw}^oFV$Zz2V3EO#B{Om
z!Lm%he$}snxO<p<KL1<}P4%nf`@kDF7)JM~?HWX|0T?SeHe#3{q&BDERyz7~=!^$a
zqz`<%$V9Kq1p5&>E%zA%ekWgd009}xIEVQja~EUHOvt*^@78Y?Wok?>XLLOOg<g10
zEjaFpoyMUQBiQkJNsQ2%Go?`fS=Yizj^G-ya2WC};~OqYg!oQ*dERl<?L}-FE!S4&
zZ;SKawYW!^u5`zHVB$~apZZgPHjq^z>OWAf;Gg^_nBkA!p*V#;S9hPrf%M={f-b&I
zAAr?;PVC+Iqe5nUv=;?PE?P0Z`P89i|DHZbtA@fEQ^vGPbOy7t6zp9UXBQuvuvdhv
z=p+0s)(-I%TV}E=9^x;dhm3Z0&F6@V>13O$Q3^}-V>|wZDiH*1Lp`(}I+mMpE#2P)
z4X5Gl0>r!(IwTMU^#(~SHd|-V2W0@lKAy#a#eNy90Wf1`M`UGFHop|rcA%SJxk@_5
zkX#=!(qFE~-mvjvjV-=@0N#SwwXfLu;RR-nuXq3Wwy<yZak22_gcgG7Jq>pyK*vNV
zQZ+BnaB|sA$n|!L-#YbG>$*el<gE?}$VU6BsqZka2D}~0{-e-}6nI(VvymA|nI{4+
zF;9C5>*vMeCDSuPYv1wu#v!U6J-BOG)YZHvbCym0Q_88>JYIOpAHA=jq1d;!Abb?z
zEl=+EAOY0Qgh|1uYEP+~&t>`KW-sg=A71jF()pd_smc23BGite2Rq9_br9~8ztcx9
zo%*%(`nOpfX|$Y*j1X$pK)myiuzdt8;$wJe%6ucfA2-qfsHYvE$IB_$TNTbi=L!YF
z&MzTWZWoaKIP@jwRf|=+d-v0ep#;FXeR_W}RB@Cmfci(uh1Q>Mf!<jIQbb%^$oBa~
z7iuOLnA4@qiN2|!F@J-|{;c{|yZ-D{K8`g1FlV0vO$}dG(YA9wi6lvT#uvk3XW@Yt
zyx)|t3H0BX#EfK1ehEae88UIX*OR8Bbf|P<EUHraXV>5kG-I9MfsSEgqNm^;j+K-7
z1v&-Vfh4wULxILvG>%jNu^s5(ZbSbOGgOCu*PAx*ehFm<3ImRi&dRuGaCi&M()6qp
z7sZZ&`HiZ_qP7#9_=OX9tS0JQRFS7gGe$Q`fuXmNK29{~xOKAgfv_1+vk|XvM%Z~)
z?Y1)Zjmq$ybPAlTEwq`BPd|SkYw1i?B0^0%AB4TF%VjjmF~TNUh`h4;rB_@^$w5BY
zQTPsf{YYzo>%y*|r9F(W*WBVu4y)SA@Q5*iSC@cQ2YUrUynYY|#@2l*l`N7|Pq?7h
z6aHUs04=Y`3B$~BSQ!?2Q0Sx}vZ&>t-tj&U`C=E)UO+8rqr;f2r-@;lPWAwx6X$Wq
z!=mj=<>y|r*J_CdhcnNo>LSMOnd695^FdCS1jk)2EW|om7rGLomr#u&3WOX1ODIIh
zkJSc&jpm()=S{N(xYmnS+gDT#^y;*GFNORs78uMY9WUNGI7e5vst$T|X~Q(v<U;R9
znd-d^jX8mjqgz{|{LX6k&ivxo04G!2pA=IdC(zMQedxO@c{6TS?8*wpaUE*9W48DV
zD~0MvB<Q|W$$7z)!G3pBLSw8~30LJ2Zr<!iJ@-B>!>vEszGQ!lR}Xm)cU#&%Sk0gi
zN;ad6wF!iNW`DKaoHqe3(|o`Iuk~=*BqmpCJ7GTkLx{_9dPUo?n+)5$+I&+K`KcM1
z6<;3zPA_nS`aT*+&GN(&q&xgHJ%{q>$^C{BV=T3PtUGDo15i^T`wCe6GX|$*37S5V
zpJ!ujD6`7ZIY{ZixsRkd2i!_`XR4?_y)G{7tV*1@ZIp{(l1dyobx4_nDh%6ep}&1f
zj01R8_XN6w$dw_IG<J7&&X<${79}UqnI#Jk1Rqg?9KYk@c+fz_I=fvF=&htTFqq8z
z!5#)O^Y-{XGu0O&$%dEPT+vJu+_V;~9Kh(jqwh|(rUXbQouJdNM>Pa{y=S%`gNf;D
z(3hO-_dX00te|W||82Uf-y16y@(d|;Py|smHpzrd6qXq%ZH@_Ku{h<ocQk5+X@Qpz
zY8=#q625akd?;UIW=}UCSP8S|@b(gLdh7?6?@p1gh0qX7`qXXqqC7UjC`%67Mul#O
z)H3EkWuTP<iykq^){iQ26Q-^cyb)AHn>ivOnTUQ{te6Y}T#2WdjVGr$0mxQ7zQxWz
zJfQ3z>Uav&gIX3Iq16bmn-<L){Auw;^?{s9bi8HGs;o%}zZO&iPnDQFyM(FDo4h89
z4Z=`+W84`F)z|Zo1NQ=AGCC~j6aN`ye`a>&OR#Ybe(Y$ZLlapu1P(D@vdd;5pR>w{
zKGbvULby`p;F<;4;SXsYC+m;-LRL$2<**TnRWcc>LpW#nQyE1|o`-{LvrwXgDKUr{
zJ%@rb$wvojG+gUO$|cVXr4STp<+++syx_7e=q<KO8+yKcxCrDTI>j3aSMVo=YDwcF
zp2P+}q=kwu1WAV_Lp;XtC^<c4#c_3;QW66noJ%d9=WPgiRRP}im(5ceNa6q>4l=zO
z6+nAb%-=s#Rj0ap;|I`KF|R|dDAg`RXv!}Vcm>VFI>rSU<opvb7zwa@5FvdL=C(W%
zl&KBPifS&aYhI&0NB7_VxlRO5$6M0B9xesWLRwek<(g`_vnIOXh~}BQG1|Z;l~@5`
zCtA5FSOelBe>_&i94*A;qABt31uFyced<?$9MlJveDjsYs;q5%YlRbqXOwcF4YK!@
zhLV&kQa@mm1)!gdI@GbfPX-P*pwatIuP*0QF*<v#Vb6MWI_%nAzJfDmuR5MeXN@H9
z8QRscwPcfebaMPOK0FEDiFYdZ1&E!LQDyqV;jn0eIMu53AvtjZ05R^h$T$;II-%t$
zL_)<iH)k+Xu-k!OB``o7xcQszNSmllWuIsK96&Q9`vFr1dp5&t!7q=YFSNGiWR#(i
zegr_wBOnWFXmG9K;s#~^M!REX^-d<s<CxSv)u~Rm0i6yK)1~sZyO_SXa2Q%Plv?2F
zWR`bd2qjM`0!5*zATlyE{bf22y6+M{3E*rL3_2S;g#8{xLa)Lf7DJ6mk-oRzG4a^Q
zHcJ2vm`w*d?V|eV#M2I14?;TiQ%>WbUoo4BNoFbL-3i`zBJN^(a{w*lvx~I<TmT9b
z6xsd)5gZtyhP6ni#XW;>vAo+yf)n3TCBYNKLD7%UVo2=b+Op;Rm}6TM*F3;IUn^|n
zL~5ZdV7U@UbnF?2Pn0=?R4Kjs(O73;XD61lniM38u}$7T)f;YNFH<%`g%gG;s+6og
zDw627lcS>W%2)5zP!MpMMYR#EsZ6~nt#KH>86SQGDgc<B5Z3w=&#T+ETe|RptP^~p
zPS}Uz_=ecd5EMiTHQ#&NRT6pD8nY(WFz9Gel}5e@+DreS7B>-@ic)xE)2Zcul%~L>
zQ^j5F-WDC3WXg;#m(t{nL25ulB(0bVgn1d?ilV_;n(|$b*~!t><;ixsE#QC!iv_R)
z2~4Dclj|mn#I?wu`o^6-YoAK4@9jl|g|KzPG7?$TzZ_ww6<$Q=FFJ3WUT<@fLaWn4
zmw1KVNgHU|?b=CewV4SET2-r~rc@sxC<(YD7=MjS64NS6xa;XAb@z27K}8^tE-9uE
zt;!br?K)SZyRIO|=qqs)j@77JBIY0|g^&er_yq3Y@x3Y+Vz)M)7pYPZP886o_``|I
zcJ4!1YElXgG!E8C+8UL^FX&a2D%V)bZgMkxzW-{S2R8;i$e<JS&bCBbmli+ohl*#}
z$CV7+-i?)@d4fKZn9!#u$rZwYSKUmwC!^g_SoY5uv(Il_s?Cs{=`r;EzJC+0-V#Q)
z^MonFBvv}enLj|K-wQk|#RqSu)Wu@}JyeqxIo|<@c}C3+wS8>p*b`zm=>5yZ(owft
zm5~7C!NAd6Q*(h9@9_*^{vb^8`-!0|{^8MHyH;1xIc=8YyScd&u)3t5<;`()`ZQfq
z^FZuj9HT0Inknwv37!_VpBF6#BqUG5ZBAv*VLYtYMRx$*=(dqr_2p=UT1u>X$xVrv
zy<Ng*$C;|OJAICAZ-k^OIsRZ@*Y6hFldip9_0?iPBWHHsI1{AMyI~n@-(XHlLN~fq
z;cg@RK)6wXi)_+4VF5&ifW8yGchZ0*!xHV?Q=V=N{T6t;R0>Zofp@P@TP@2_VaSSy
zBkYb_>I7P!X%j2fG=Fg~y6!Cd0_@dbf+^*}R)-c1VhIGkY=k^fhg-o3zH{zC4OYm=
z5&>7+E2`ahTW2MoLi3_IK`!2H|F*9(gXXVWD2?nQbf$3eSV1?xC52N+&ZTyB<Ytta
z7BPU7O`62jfyrY>o=?Pz2fY0JFRt#l_RvC~WAjg;4)_o_*>!D#Psk&SPSws0LpYC9
z`+G%!fX>6wYgI2Q^k-^>1GY0<hEFN|kZ_1y#@r~>D@AeW+K5(>QqoDDdZ6NJI0C7W
zdLn|-gZJAYbdT9(o9oTjpks?Ig%R03AkSmmtf>vUdMmfc;IRN+G(GPy7Rq~|3!czA
z#JS3Fl+!|m8wQ|K7534xdlO9uOI+}5#|D^oro9z`T_~2#WT}>Q`D+FOCK~pbCPu5?
zk?#F|Pg7=92(7g#8b^;Sv|6)dk*#i!>b|x)k`w<92bh@I<C^76pYn@;5rfG6r{#-#
zS<yif>of05aD?H54WM68aKv&wdNHgoLjzV5-K0{&mQF&M(gs1b$Zptf<A-U7$0_!!
z+18ab)NR@fD@K64UyF4mHa);(asqQG{ny9m%Xa?CCv<CkoWDkY;WB3>V+SgP)SXbQ
zooJB3arrv&Uq4nvjV>Z^tw}<Aq0-5svx5W*@saOHUlwM)D|TXmcb@^mMXzS%^@LCY
z{;|%9a2>*uK8ijYdb+f}{81}YbCKz{>WDWy1S`2X(s26{x5SLC01K3x&?AJ{q+nT#
z?%bTg)-On8VQ9{VyB#=ssWsGnBef5LHWOBiqR)$_+5;Vp?8r5VscYRP9UbEQ$xM&I
zX@H}+8&(Uj_wOtqR7e@@(YM+}lfw`l1_A!cy|?vmTd}dp(<xz$XE*xhf>aRFrDgd{
zwn6|(bd8<PiIAaQbx74bcq}t?(CSdwG}(LLd5^>XJ3upzn<N;Y>*DXQ9&~HDkRVnN
zgO43#2cN=5Di}Q%et}tBtCFgdDX~R3jVjd%)nHYJvhBi?&F<38TorEbJIk(<sK3~^
zuv#TV>0~i656DW}Vz|iE+1Dy@xPmdl2Rr}*q7L;n8n)hz>+><MxXPX-y%eMK4Mug>
z?}mVu<_*^9+}CI0^V;4@E2()cmWX1-3|^*8^>u>5k8k#HzU-9TkFZhG*iNfq(yZsy
z=)FykDM)c8^+`5%KS2_tNr|eAl**x{`AY{L96tMUMMg|p88!(fY^BgYgEn6a&ueQM
z@LUZVr7!9-R<u?nDZ}3glF9E<m-L2!YF{FYCb%Xnp28@5q!G5*2lk(K@C|Wv&DgSI
zX3o1GFSXv)Y!l1__>BP!Fn7u%yL3~y*}}Is(-N^XDR~&}ntZ2L<Q&ouQF*(I#c$%p
zy+h^qXA}~JLZ1$)=b-P6u`F5Qm|SF4riXt$7r^sY$U1w$b(J?=7?QH7b0|}>MgO>+
zE`wd8U{+?;i&v;&&~|3IZ31%GDq9RRA-))QemL$GGk?o6i3X29|H=*DXK)lg*RKq@
z*D@<(25GtRv2}xXSv%NrlP*N0Gd_$%D-^*bmx%f0X=n~uQVpDVm>btr7S`7!GsCS%
zA`moIm(h3{-iLxWNRCgt)p5YE&b}D_X{Arh2~lTlOwpYZ0JNWgn62>(64V@Hsl$eH
z?krX#Kc2MQ2f~_sYo^Gv-}{SNP2|O~(LCQDFY`c2oqtRJ2zAH#5$-3g7KFnM`vTeQ
zD`^Kpb8_)}8d$O3&h6s}R#+^N?eaLkddxzQL5<hdxUqS)(>A38>R=q$7*;eJZm)Eo
z+8*2!_0fWB1S&syw~%L7CSDWvcSKo7TO`5dkyDDtJ-IOoQVwh<&88OkR*d7#98P>#
zpDX4EZle^?=g2T=2+2^_xA!6cZ@qsB0jxwXQa~<BGTkfsrN1~o5zg!2ywMkM02?Ww
zx%ftszcCuJ`#Wu(ke_Jxu6&9mj~=z?jdCt89;s5Z#^SkIwxOXA_K24)P1|8J>NigC
z=4~-Ifpq#<$Nn{lh~(HED39vvMY(FS&|&UZn3>yO#Z@m{JCoQC_kM;SEFuVw9-1_R
z01B;LJ`Culu=!H@JtwaAd&&M0aoA?vkS)>+m>I9|(WZN=!?*_o8_m=0Mx>tm!I)5_
zcPRKeGko6<o#XUY?4Q^i`7v0B7cx=#+}R^A6Z%gqiK8(=&If%+nG2}TuQs$J^-IJS
zq<v+;ea~{9he6LRKD*_)@^md~mjI(-X?CJ+=|kR<{%4MzR+jDav#{3QIc(Hdgd{O|
zd5;0B>b_krG5bbP{8MIcXTf)IJ_n6R{WmyD5Sxf68VoNswegBs@;?K?T~lh9azs{P
z-!k0+!v)p>49nLIe*whuAD-*v_lG~Js6O-<h)*l`4sNRip3sv^_f_bDSfDS_Gu4`>
zc-06{m0u^soXOqlk>HSV(}SfVaAFkr8lWfDw4c5@A7q{Wl)81*S%Kc$BOkhMw9GQn
zp*M8{J8CQj*zv+*ji}mo2nzzen@iwh>rDFk7NXT@9NAzD<Oy^FY}P$S#rx5uy`EcN
z@&3+FZxSqGJ7o7lZ)nj5*`l2=L((<q@I~peB`80Fhde;M){-O75I&AK5Bs@s#l<bh
ziXJv@AtAsEk8FLNP()7C)9)gQB86C?oDACpl@ccPdI?(Z;k#+7=*Do<sP%F7Ayzd$
zznVT*EfV9FaoP&LW>(<L=Grf_D)osVp>N*1A=Z`S2}!qhQ}H2e&6WnsG6Hr=6H^wl
zMiOYwgh#MhfTo0Y;X9GKM>LeFL5=!UIW_3J(9GMrMU<$Q_c3h<7`v`ut^>^>q-k|T
zY*dkJ+sB9X;*9kgNd7$M9|l78z~Pk9jTw;PrXVeVClRWZRzng3q=EQ&5X3^{lYWiA
zIH46#Uf)8D5Mo!eX;?_7Vut{nH|BAK25n={E*NI#l!77V&T_r>N>J<8d%hrkFg#%y
zqwPW<?R!X*q@^Ae<sWfb%^=f2T(}H%o0ZZS9s)Nm30lZFdz$HTS~_Y!34B8yzX-^n
zu<nE9yBF^%7!1VPqtlV2e$?_Jrxx)3(q9OOoax3$Uf%)hG9|=qFLNT$8PdY*lQ@3|
z8M_U4rTdm4lUDfb2{eFp(=o3|J0%6?*0w^i>q3JfPg&edoFyO4q5r(qoqAln8lL?;
zA;IrUo8S4S)1uVwH-{<?B`lr+!}51mD99Go(6Ki|N8tv*@wXHi062+0E}g|Hx7&aO
zr#5_LO?>1>Qo=*VYTI1ij~?x8$4Gk^u=-;+Zg9J+6ity)=Xs(JfbmkrkAqRFXg{dk
zw+?LQfIl}(_4!bm<P8N()2B+`l*`n!`w8ZN8eQd$VrY_}6-CX`lgWb?fo+7r(PaIi
zfh70w2A}0qNGCk$hWDvr_&jzU3mbA0Hc%YLw|gcy@@Q0GSOoNB^sf2pfYbOBCk@J%
zLoeJ}?Z=o00$_<zsEHcVC3)2%_l^h@Uh68WNdXSn-5K*Ct%>ZCx336juEVY5nEF5S
zk}y5^K7E&_bQzjF&2)CYy@3XkJf1xLB4r2;$vSQ63c;af^u8l%V2>4I3sYaLuuWtV
zQxj+oI2dQK6}NiiaSx~*llMyh^(5R5RUsJzm9<*E-3P=L&{Z<(F)$f4Ts;~OUcrm>
z!dR;AX`|+N3S>gM)cQj)uRxaMA>QE0Dq##2mGzh)ZqOd;a9^f#svK1$^V9pnS}HlD
zKmqyTCIOsjzJnKA?3VMQ6)st21Q^JZn4kA~C6Pt@#(AAb&MeZrv!BA}9n0lg{`!?b
zOcWqsYlrukP*H#)2X6(z0XnFZjyFCCd7t)PQ+3ZGc?cbqPe+O-kR|rB{x{_qm4@dm
zIb}7K?=tzlgA<wMxS1#Vjf?k9PXxXC)t*$7(H*7I!&@<UGK0l3Q*dhMPEH@}i$1J@
zH*Zsb>toaaA*-uPiwFXIbfv(HLaEfSNp9wlr#(J9u6d``@JPA%i$2kXuBpGsIXewr
zh{5UASNV)TF~Xz!s*sU+00BMQc*ZC2Z;mMl<Dd8E#hSJ{iuGK;`r3^W3pajir{YH<
zolIM@io=u^F!j#yr`+t(V>y*A7I14p&VH-#LvEvqwbVWQ6j0W@NO3PR(8d4}_9yr<
z0yS83$6xJqKJ8s()L34Yw<%7)f)aJ(&vm?1a>i$fs@y_Bv>vIJ>uxbwR;>01Dk1_K
zP-ffoX{3&TD?~*GGL0;~&_t%K*VH+FQk^M(3h)?LQ?;*c7bZ?BfyZd{b|e!Xf|_))
zC<PTT#UDx!qtgKx_kv|6qW1<B^h$~jw+c|eq42{gZ~J*bouoa{*O}<syN5pYld!IG
z!nIB#4~Bt*z;r#WE_}V*UuDR@=eDSfh!;AynxEjV_*~<8*(T~rxo(t!Q^%xuW)Yio
zu(R#xkzLBaeCCx8rdw;>XG|_l-Ux`PeCs?zGX5PbA^?=;ACFSMKMk|$*Jq-cYke8A
zTG!3Jeu*YprYU6Uf$gyMljQ^nDpou<#NMr~@Fd_?Q65c1KSS!sBPU+rg&#}!Y%5iN
zll&=zNm9=1TVwqE8(v7?eNKI?&8Txaxbg`+oOBI>17s?TkT><bL!%7-dIV>UqK7WA
z4W{H5=C(>ByC^xUI_)F}9c)!E6`hfU#tFvbajt7b|3MS(lO<jOe$k;j+=l^}!-GFV
zmfwLLJ<0g!*nu2<qI0qYW^4E6LZFH!o@|w|wdUjMpPprm1{<aJb39RRMA9SACQ@0{
zyTrw6w&n`2t$r_`g35XxE8K(dZXFs<!5ob<_tX(De8ch=@e!D9?|v^{=Fv>3kg`Im
z1OR4Y3-EhcrK>@p8|C(d<bVjFz7>4Tt|<_#_0;Q9bH$NDTpDrMkZ6JgFBrRI7*5Ie
z;Oo2>)DN9}{1lWUOnYTQldH;m#=P0<&XuLr(LarF@!@l<AsfMB^6cG149jgn@ds28
zRKnLQ3(19Wx{EX?hErO~njazX3s;H#yBxRC0x!V@i}!SBF7>>yB!AU9Hv#!Gu{q>)
z5UWJ33AkJuYDYtajv(P$pZj()83~_cHm01cWOF(B-FyLpUq)$HSeQUXh9IcGD7Kjw
z+ris_*fk;oB&6_Mmk{QdZ}6hSD%=h|*ya{jv7|AgAQET%WA11rOk{v^_SR+-$Xgt*
z%Muqka(;IejtDowOg<JQjDC6sQO<4#o!j1Bujr*x=*zFAeY^wSe<LpEjI>{$<nMq_
zUfeFEhulW;kgvKxLn$cLw4c|w6X!8uJ~0|kX7d1#Dy)DYaO+xtC$sWYnkW~Js8KF;
zzmauo<5l49&mZCsQUxQ1v>d~{Am=r%mOD4iID;&45k+*<1z&020xRLq0M;sq<w)_c
zS9DjQxuqMidz9(gLDn+Nt)$h=?+~`LtB<Aw53L&%RQii96!`fdDa`U-HB&XSfx7NS
zVEdtshW*vFq<^)u;=v?mvRm$Ups1>AdZx*2G5EIXt*HN7YoM|JitTu5LJv^lAA%jZ
z#hYySEU_tWwUBANnQ{5v;$@mHOcus4k8KjfYU>-@V~2z2dYCtSmuk-{#t<c6HE}ff
z5YnQO*8OZz@B2EVk_IAlb`b;_YDG??>oG3|Vx`QgrR;d2HwXUz1)m~d-49WU1hdtH
z%6hEw(|cl_39*n&A$zJQmVRSMZZ5CW3vE<UWBNmR?*Ypum&?BJPE0i12xddCK8kb<
zS|MX0goquV50nj(u`rsG-6^DPwDQA$Z^3ie<Euy5vec3doFT8>3%FXrbGO}q#lKwf
zVxNeog@5R=UTu)2E*0L0bOFs*F(oV)3(mly1ysJ%y7vL_a*;mz|HnBJp1_=`uFp$v
zq_jP19lv0`ndt$;%QUfyLsG`WGmNpbY&98kn*Lc@h|ipo3RS0ScH=o2_0#}vwspYn
zV)eSWdEx&pP9=pZb`NsYvh{qMDvF>{S%znFXztqcW_<;p+<ptqLVl!r6N?d1$+{qw
zgTdGv_0aJo_x~EF<hFRgF>Hcnoj3uo5ofTzr6q}$=#Fx*L$*#=h_<W^lmj{l4S!iP
zGFIfW)K3H(7H-w`q+-egMLTqvaUPHj#@w0VC|Majngh5q7`N}H)mv2k{f5vT#M0H|
zHP~{*4tv!;#2F;#`m{@H5ob`qOaP3nr|TxZ$t$_|&hZg{d-xS-g6m0hC1Zvq)7l3_
z$6$E-;kRjn9_uJe0W;P%mOHYVUDu38&kf<l7<!8m?ChSZ`<wZMz>p%Q==Y&02BY-I
zapwqmye>+GZBJ@1?A$0&7X#zhEkr09z<a!t>{L%a76)29f^<^U3%)O*OTn4m?HNTu
z4xgKIKcdjPD2$dxiPXnuA2vqi6uF3sA^f$&p2mZ22aRiurkR#~97vJD(~%$?l(10s
z;v4Tj|6mo0W~dema}KCt*Y6%2&NGCKovY}aJ0=vC)g_=^@>L|L^CY=!jjOX0D?W3b
zb{%p>HWmr8;mU!QGr*1b`U?bl{YE!b)_f*H{@4mD4Je$2eqhTUWM-*N!-i7z&S^jJ
zWnH^gTXn)V1S4DDcad%A*!K9qhI7cgtK0Rm>JIw%nHYQDJ0+0mwxXth^_ZQy4-fSJ
zNC7WxLP=XmyHQpETVMOir8yuFQTP_1gfF*B<)yfI?)hMF2qhx1D6vrBi_KJF-^K)$
z-lkBvGD<t&T~)5W>?|1nRO&?(p1q@RISzLg4&)B^7;{2h|B&^eRX*5jS#ifrUjX8a
zol}!ATG(92wr$&f$F^<Twr$(CZQHhOd%ly(Wm1*g?Z2@5>0S#mYj=u0xq@yiKFUxZ
z){GkuYoFHUVO*dL)sCTNLE#vN?U8cMBTJnfP-L>wG=N%*NFZ^ObH7uBy0|JxclTNs
zuAigQ2f(?ATxYE)>hv-A@ZdfgxoH?6M=>Fv_%eWG-o)4j-u<gkH9-PzLWbT?wc_W6
zD!#JQ$1eMAU^(M<3^o)RxF{8=Ns+*u8pep&^1~vI-oeC;J>lvyCI7|6&NhJd%S49`
z30Xg}+!W7l60eGnOWKDVd~+Nj6pAdwAh4uXlR+2jI)cUYpD5z(t*Y9`Ru?>=T4NMg
zqP`v<Tq81rlh_xuN2={^`UVFjEx&a$y7#K?WPc4GVpCmovtYi==RIqyBEDBT^dIw2
zas`gOH|ni6TKxeK*h)#by%a!P{Tf-f$~Tj`mtot3xcZSKN?r^c;WkvAfTh|(n)VVY
z9Xgga>LlAA_?N#Hp|8a|hEQxJDaxUWbr<*)Ue~tG`Y$cW#vxtgB#jR$iQ)u>+Y(db
z0>qtNkHC!%IPbOF)-T_dlR!N6>nds8Ak~WpYVV7?HGx`*5!=?O;3op(;EUMP*+bqA
z{eXjA5o6$&%&oI5R)4TJs%8Wb=460euVC1{+gN!sHJ;O?HRSWUKY><OiuL*7v6n`b
zC9+3BU8McUn&;;U<aH*iJJH*?Hv&gub!<{Rh(h9xUSA}Nu~i-&*fdvugMh!DKl#_b
zmkw|l&Y+0O_=RxZHLV~YeIRFnQ0M{%Y4dh!#w^(JC2~5?m*t9@Id-83X5ogDd0d-t
z<=8aG)NLOZnnBsepCB&uYwT#fT5E62U-JWz9LX(@qn{FCaS63@l_>V-qxr0`%*S0G
z416AWdekQtzMPfW=g1AY2|GJ<9Xn^f;>`yYsBA+_TR$hykz??(SvJ~BGyOtl>v}2x
zxoK%%%DY^S$GNKYlmQH!W?Bg;M*8v6jJCGaM3IphjYVKC%&eQcFS9vO!lUB+jSk7s
z$95Y6H=o*n{0kA(eP`{|86_@DLGW3H_Z$g9r&2{f3J7%sTUEXGWHt;%B-YjE9M(!q
zGRA9Qmvr%gbH;>d+&qE^S*Uk##%Lijquu(Ebu5F5{va|t$OmINp%xB<kT0En03*eU
zuJ-&yHWHwTWGnQ{`RWD|fVy?#ffy7eHww%NAm1Z3?aniT8j1Xf1b&@bwb-2Qbi7tv
z7`P2lf8fqtaVvowLAs0XFbH_)J4U+%^T%G;eW3mjjkF!v9rmd1Z0(zynxU#Y<l5)E
zI!?t5Jz-98`L?{La4^G;iru6(eUp9vk-QUS5gt(BzGxxKyIw082N2yuU9`ujFoNH#
z%-Z6Em$XQf`7#l*G<tS+FBo2gLE`yNBD$q$4om;VjK1f(`ANC$^il`mDt8u{zn!sz
zr%x$tp03TRd-A23LTr~jZ*RL}8e~g>$_X_me!|iT>7gPK^0K3Q<W+MV&K2j0{H*mz
zIS!_jX7WL^=8p-;IvpD3Cq$X+bQvBh<&G(W94J`|$x96^L&)>PweV-J!zpoP=J5x8
z7Y`$u)CBu`*B~Unl^TR5HN)~#xJXoa-+~B-V?`OS+jHfxg~tInksSf}0o)i4(yh4b
z(~U3$h3ga70=(fgR`aoKx+v!AYAh6IDE9RDfxbP{q@m9>A9K++n{)CJIy@ZZ4>tPz
zMa}MSCZRbyAOhIU-q2KuwSPiQtoe?0q<E&w{*&B8#H|0In_AHTR&f5Vmr|)ulCLX{
zqJH!{GovBoWl_ayGEdF6gI+f{+#^Iuz+YT_S+{m&pO}K$ywVsXO+rn5{yg23m8pjU
z1jSPKiauU9?IW5CB`0su62}}w^AHn9nE$~>kQsUNAG3w7jj#SUwM)pad0cuWB78(3
z$#D|sBAvcsh9l3B`z1^;oQ5!I+L^n)Q?3igjx>gS-IHg5Kxmpvm*lMw$pDVf5OZ#Y
zF}*F=;3?QP!DXn0m@D4$ij0v%UEPp)^VC`xbxY*Bs6*YSw$BZCzpT<K;%ZM69av2w
zu+fJ?M6fA9G98BbxLE4hN!b$=K$)I#W-RRanx1yC1D4??LY(i-z`0CBL`xTj@y=aA
z9;1bQ{V{x!pRJV3&S3x!5z#~z<AQ-B*~rb`lw|$jg!Y+lp%ZlQ$=-1zR$keUu2sv6
zo#lku7ii<Rv5^Eo-e_{zE^6EhDLS=~bJa}P*Bpw0@sx!e{cNvToRWW{|1S3govgn<
zHgs=L*xEq3;IlCUEf2uq#bxuUMTlYObUosOyb*LHybxIa1n3O9HKyDBAsp(52UxU}
zhka7CZ;RNk8}@DG8kBE`9;~1Ni}j>}n6T&TMs}aT7=3VCQfM?)p)og*G8~oPPn*_`
z-ssN_dYBPo#+iZV1|<$UO`NVkr&PhRuF{Q~N33`5=xn2VYrSDJbq4c`@_<;|Gf2;T
z;**RYnSO<S0bIk5@8`bWUG_(9&hIjwy-x!PqmnI|%y%zqn0Kq4Y_I0pjB-El@>P>F
zs(T0SW}oo2t9=_IYe`hoC>OaB%7A#%;0gAIsJMiAib^j&|I_L!GrX0wF1_l@FECp`
zv4vVGZ>4=QNnjM42_yuDb>gaQao^~Kv+w`msYFyJmBBOF6??m=%b{p^&Mo95FPtyH
zm}4kwq|Ej!fN%~8<zjCy?p=hC#4l^Iq*JoS?Wtl=bW`6<Z=K;3tAy8q$kr2Mm@H&2
z{y~vghDdo_Uq?Sahl)9vuJEBOO$N<L2(dZ7QoD=6Uu47=^%D$_%}GyH%+D?BT@7D3
zZ<<A$H+!fQpKIBd3kqcbj51WXxzL$Ddww;Vp2B6wkBk*>BjpQ+&C{L8F?)HF=d-JT
zQrwcjJ1jY9I1`(lcRlI0WE-PISxML^Fe#2&-ZG^k6?$DOTt62$B|#C^$s&Ce8b~Qb
z1(8eK)2^$-tfdh1h#-KU<zzTD3~ACFF~%%BF(}N%GZ+%xh23Fik9JkGZ%D8Rq_41}
z{l~QQ9<c+9n)oS@12bV(nGnO(3i}`Fr$=vymC(}o@AIgTP<&PVM<yNuz5c?L4DrOZ
z$`W|l0Y^p}zL+`i1w{e~dgC7uCvANS-UyWXa|vZ*Wf%Qu%LvM@@_xR@yG=T``?NV6
z@zLlN^jc5An%alnnut<axV<dMf%$nu!6g>Xh*#pM&5|!P=CZ{AImKo!Fj?=9{$6LK
zZ3b}9zRC`?7t!wFI{r&%T=+atU*wdcdF8IEKQ$0!#y?dJ%nrP=6SuZsxk6>`;~h1X
z)BmoaJ6|TBf5)3X5c0-ajn!X6=Nzub%xrN_JLTMOH+}!nb0j)Wfz7orKv+7cGc#<v
zpd|lW7%<JZGxW+igl}(#UoybavyRy}vnoI4&#XIdA-!6^^lt?LyM}i2xncUtty%}i
z*jrI*3*s`-)c`w;BJfa!PP+v>7ikzdn}3tKxMD|$=eUw1=qsx7+Ro{6717{?h+q%N
zJcg}>O8!zy;`iU6%E1pfPUK{v9ZS&nOdI*!4dG((j)1o}(6_ycDNMFauwuTmsyVrc
zR-x|CrDjQ7m&~~iTnBGx%hH&wq3S};yS#^OIJh8|V|a0nM?#czbv$OVek$K;v#!_g
zHD6%(&@eaT3l8I`7b>-fMu{-tzQ2eJi&!^n_VI?w8ySM_Fj6la)+>2upQBf7m(XiZ
z=N!A|TFW92!tKR1!<#&hs@7Q>=K<=gru{vB<O(7gWNJC@Xx;0*W~{F@ZWgor8dW*}
zvJCK_`#_it>IN)|q{xuqdqZN&{vp^&Y}Yl{6NuiN_BRu&=QLmC>8yBeH>!x`fQuOP
z06HVrg6K&1tz!SY{V^=x_BbP!&ZRw^$4&u~j8rF@CAm(KRKbc7PKB~gGqKQGe(*xe
zVr^;ln%;kdrP6InaiTa=cTVx%D2;CP_6Lru0_$LT{YU;%u$kEb87}UQGZf{9CKsyE
zzhPag2036`;e|xqA}&K>PENd@vG#fUs+7UI7b~3KzZiOgO+J_O`gRq{q`z?B-v*_u
zjV9p;oh4277hycqy99d-Uo<-o&?r`r`-1a93w6&FyBZ^C`<J71EL#PP3-3c+FM8~t
z7hEd01(m|7w5-i1$9otW0OY>5cydwb5!n-m?z~Rz!<54^Itam7t}85&E>B0d8D5c^
znw@4|dM^@O4n*szTl5abPBtSSH?bM*Mb&4F_fe@s6~F+`FwT>`rlI`qIQFnxiXmF$
zFyAUkO=Yd3%qI6rhy#Iv5Z53@C1#cD8QwH{Z^&xZn!MFTL(Q?96h$QvF>{o5{TYjI
zm|^Sa^<k{1y^8e#2frTrg@keFP(C{clZC5Hc+gvf(Of#U(xD0d<1CFS(`t#&|A1}k
z%(~I-xz%UO)x^yqUFr$na?UC}2;fym<K~b_P)$FGh2}N*y^VU_Vq}TB6s1Kh335SL
zQ4UNFMx|T39@kCDY^s~0yrNS}iWaF8o9wr4d=>_GmM9D(AUbw<lfp<F+#3}Sp@6X~
zze8lxn^hFl1m}(}oCsYrf8<;FxAX0jkA5@5^<^|6lup5NdUPxbRCz?I=hBu~n2jLW
z2=6XqZ2K6migH{RXIezUvPWRsPj!GL9r~|3<*&h+rQpx9Fka_zU#FgZgX*zem#lfI
zu-gKGSZMD(nY?9d(iOy#u#(h$s!<~$xPWMoG_}~!enPDjIUo)Zw+1<|x6Vgmr&f~W
z&IN}-Ib)5xjtC7=B+x{tV8nD-3}tB<U6l|7+mOTiUeG~&_sgb~kRl>ma26y=Z1~g(
z8{$E(1U%$lFkZ9#4d8y~^{F6{`P_v^%zNmi1anMfI_i6A2i!`65*4B-A&pN)+kRh2
zZgr~`Yg^A2q8dya5VF%V04N>$1F*jgR@A#muy9};-LBDpxc{+DmE_ubsGS$(Rt1L%
zma~ThpVJhM<z{6~==IA?XmgtZ3$hNp%QLUEwdT_2d(J(i?B<~M>0x8tE}|Wa-0q=1
zx{RkJ%>&ZDT2|l`qfW(WI0{D0-?Ou0?REjonFDqni!q@f5M9;{A2iW3Rl-IJavdJ_
zexNil!O`*7cHbWvI@dBb>v#Bj8Bh&OnS)_|bEPk~_Pvrc3nfPtOiHlTu5!AqoUg$>
zu9Cn?EoG^2<I1dTj}_)@EcfaN*tQJ)yj&Lb)nc0?^07Y0575NM*te~ZBm=D|#7?BW
zqVbB<sxuV<kTQj#o3`>I$}XUv%y&#dpP`)G)q9JpX$OZ<ettL^h}-tA6m-*CocE~F
zXdnDA@bG3T(##0eh|Dr;e0<;X>O2{`$Jx-%yo1Qi^go756-Z7(_BKq}*5t}-R!rt!
z&*9o)j8Rv+<^BL>0}%eq+J3ozYb`UhkQd~9o^wY9HmD$J07qya;}-Va%r8d6QoGf_
z)q0dcu}(C6{@8w?3f#z%Xb);ZI<ukQ6fhys+H_e;RK*zsFvlnD6LDop=ip|gBYQ2q
zftsHrsm;ZvfwabR4^uh&E=M~#;G-<U{KJ%@&K=pk9*J9iYUxWLNEYM`LTE2UGiHr9
zA+WsuhZ{@HOoji@5Or6XWG8t>OI@;~vW3DwJ7dAJ0<c({>(?_ZN~N*KGXl1Eo0k@%
z9m~d2`!5`(J7BEVS8ncPg;ZYZB)kcbhYb>Tw<VC!Ly-z`sfY@;EM+S2cwn?EaJ2a1
z%sMW{q)FY$NFvo~`yiD$VjzBNgs+N0AW<URHVHGZZ^Og1wEzhpgy3j?vEuuAB_|~M
z9l$6Zna?wF0#}I17aL=i&CQQ2p9RbYgE?T&GEwN#tLZ)qd*U2A_*|ktx{@roew!|*
z>D<LFqQyjTr6m{K2RTw)rvk5Px)n8|Sb-ief7rhuzG+mb5pm&gRwQLlke@d7=kHWV
zJgXf29x7XeKo|5iNsaZ0uxl=&EuE2tkbD*RyL}EJ9dVy-+BFoI3ze=^VY;(d0)?E7
z=Ee=}=a>fnc-J)`t~^P6B=oHVe==)X!;>-iS|o6jR9#wi#koV`d$cM^G8ee?@w>L1
z;KOw8r_0AkD<s;U<1Y?5+*Gb>8{0Z+^!AhL!1Nvh2xS<li9Ls!Nl}K3-b8s_<xI8=
zYA3oJ{*Ze(Zp8pGssN$Bl1F<9mlpe|L&(YgYpU(|SL#?OR42%~vH$IPeTyRFnRrxQ
z#SJ*mJ?aTX4Rsc4cG{nre$hw;g~MCWxZdjd*egjX!R*_jCb&%IH}pD0GwpSd9YEQF
znWZBd8?FQD2%|_mb__3U13d04xEQ0rU~-~@4NF)ZJkj?Q*H~${b?$Rnu1O|_t(Lwl
zfpL5XdU_-8c;&0!BKP1QG*eSw(O>IHi-ip8{m77KHh1PX@I%pUW2Y~WoqMmk%Qd|U
zw|!DJhkYrw*{OAcMAWT{@wH3=U@CO=awgA;F-L>lu5OPqy-BIbyjru#W1;;WvzMMF
z(Usu}-m+vOV+t7mw&x?wRoqBYWm~qk_NNVa=6qrWqlW%lcsZhZ8MGab*h2(Be)HWP
zZ3eo;X}$t*x-2J<4#0_@{95{qCunNe+5^dhv=StwDgAA9Ic8VRer{i>4A6&m0S?rq
zZ60u@z3UPMA(YRnhc>@!n$s&u8Vg>O$O#XVjHJ#pp9O^tWfw0}qR7Msc&n1k%PsvH
zEgyGp(*@^Q`Ta)Z+|+7)``{W<T1G;y66h|BkkW=-yQ`eA$eU3W7be!mtl*Zx(%?q$
zR{M>N8;!4YkKrbjdP~f+9V@`fGced5XdBGcUj66<ZpBkD!x?J6;G+3>1m$j`K=7e>
z9jUuIr@nqNPrC<QmJ$<?be~Dc@jeONRB^v`svm?TDYj9z;k&Y%S8Wh||Ke_QJ)8|>
zRfcRJi*>P*ufpyH06^H`rRLMbgWj0(Qyy7n0f-hm^CnLh08`7>i0Zq9s-ZOYi!cqy
z1rr$>Qx3nTs@ckMD<eoeW<c|0Z)|LwUQa(Mw-2?58G>3l%g{STb3^BaV?<Ks#b<{n
zWuR_11m;VTi$?oMd<W4P+W3wkm}la;o5~&hfC?9>L8q!i--3n4lzcE3z@>p1(h})D
zWTN)%^OK3&FLOC4NXoXM`wgh5G7c?|ir?o>TPH)b`bcqXq=A&HV0<X5TnUs6C_kK0
zT`Nq|4X(kb5g=eu2l#oKK6Y0~CHGgJlbeqf2E4x-#oZlnA;>3*n|z2|`dWU;)0;+=
z05cQyqZJ^<`i0NSTS&QRMjX00!RS#7s4T2EJa(GdtF;#V70HX5S$$7ZHs1!>Z<6{q
zv!r*kBvMN|=TD1t9QA8%QOSPNpN#aGN%d+Q5Fl45Q%BwOX)e^-7D`k~;iwa(xj7Hk
z39jovf;*Z_HuOAIk>LheRnfzIM=^pvarqsHUil48_EM--p3G+%J`f$(&!n~H+aeSO
z2wA1wP%-K9USHkON%FSHWhFml0@ZBllKXJLmokxpestvW=37edt@9{_G=H`w<C!;o
z+VJRhc6*yxsPM!I?BIyQJ1$b7piqgZ-i<({_;PpNrh}J4^*}ERbVypd{!eQN($Rw<
zSVWfTbc@_h!B%$8CpQ7HzSJ@`JU_Ibf5GpNe&bAv9Wp+=eq$S+cj=R@rP6_!aP?dG
zLOHOO8Y6?_CojlLr()Y|#$X`UqcS?ARa|ETRo4xYxUJ5X2{I46Lw4)~#JLX;Jcv^x
z!|$dZ>I$f7SYU?*bskTmn8OH&Uqt$Qy5d-GQ#=0Yk*{<t9kBVDMgDAkB2S~yEn2Nc
ziVQaUp%r)zCIFcaot~x3JE6nmbUx@l$YiQ>Z`~F-y6qvr>z0Ql5osXA1}4n5mAX)A
zbY>Q3WVv*iGaASttrDKmUM$kA7E}Y&<kBZCZ$rn<w;oh=yNZ>|#I_B-eE@DwWH|Wz
zpZakj4iZW<LIRcdY%ksY!~LmG@9cHW);y}iMY*-Tgg78hL@->zi{*BJsV#PAS9p`G
z5vzyn3v{v&ycFN`N+dd5VL+PNcZNfrcF_A|{7_hlPD>!|{E?1tbduOHqu`d&6=ZXG
zQt_;UKFcd|B~z|<|LaclYwn(cKMV7%svWWQ$~DGHmdOfRO^aLqnF*+a3NBmD#VxkT
z1U#OOdz(CPo5mOIuOkOct|sp0${5Xk^6-aZpWS3c*G4ji0qcyVTSrrdB8znker}2#
zVj<lc)VS?49BN?NE$!Dq6IUW=XD^hceTj-|Z(jdfsb~uXL1e=7cNDYM8@QcjF|o(o
z5G)X-u_b@;h}!g1e};@XQO<_7h}>$}L~BimlwpF_>8i*}eE(DsEy!=(TK-!A6iyGE
z$rWZ~wxdBx*ht)iip)-l<1C|8<YO5MN}cd`tQ_;YqYj>ye~bIls8W6~&3)JT(o@#)
zc0WL=VnOJd6s6^F(ICqolvJ@Cn<#VWYJj#DUT8fO6l+OX)qcwv+T&H(Mz}%)43TCz
zbj%_x4@zEG+0#UX8R5Z!MGM*zgv{oPtQPK|#R*f@R($OlvxK2U1#>c6!@+~perk7p
z(&cti&4A(MSAgfw+hU>N9_ObgF_kMadh{d-EfqcMDv-)?99=eQZ#eDxNqk{;m!adU
z$MDqUOFsBmAncc$oAzGTF&ruZwh1`|_pxG!JWBMamv4ck7Rz0Z531uZ##d+CBFc|*
z@S#RIw$Df+v8dyEJ{Yb_H3~z;-l|cUL>UYxYbnE-3_$}%m~(tm?0r(t&j^EY;9Ib&
z3@g6r!wQWiNB*V6UnkX_p`&OgBs(d@SPbbvnnwwE@E%sp^6^$2JfdW65G^~}_s_nP
z&KX*yNAFqLJTEB3bD;2gSJ<v;9jqJwnVygjEe0e>Fo1Fx@!9AI?U@X#3%@{pIKCRe
z!H}`|(}K5KC*Lki&Zn@Vg{1ZokacLDiEmzX;#;T?Iz@4-QiSq!&8#L-YeXI?i^bg4
zi1ORJpM}C)zXEl51^H}ii#OQQy4;ZwZ3>Mdv5?5etB?uUoShL?4YP;L<ACJ6?!_8<
zqfdsCu@|o8tjJ$V7hpls23r@sLVht%Jkp>0ly@w*h7mTm)f;?H7c1YCZYfb6btk=V
zqBl`$SxoSyL;q5dc?#vI-qts<{U3U0&vtpUt;!4ufoXLHKu5_hxJ`v#_MLYUNWaTR
zam9I&deT4Ag=^F>n~R(^{{mKmfK#Y$h?-$tmC=P`-AZ7O0XCqUJ9n<tFqa{%RE`+-
zA-+OhFIr53wpzT5L(xYoUOhnh<0PfdY>G3FT7klLh0FE4M+;G0x9%0QHfBYjC!^Jy
zNR5<rOb}F!GL)V^gX1+t>F5jofANt+l$hl&%ML-fPV5Z8(84>;v`ZX!lKNF}tI13C
zi83qYCA0|=QIco@d2X5-MD8l!8AUU;1eUGaci{b~lf%JQWT`kVX2MXtqqYHl_+6-(
zI~D6prs%VQ#gwW6XUhrQZ#8O(wG|Z<=7KLo`QWyUrM*+o>q|ergZzx+`WfLrbyoPg
zK~|yBH(j3*)E6{&B)xa=7oRj*$%@et8V?g!qEg8OJ1b_I?YbOj<bQ;}FkxYoaKBrR
zkbtVI!<<|$mo*^{ie!zg-lFf?Vsu^7_g7a`3T~MTSZ23`9j_Cmd_;Sse3~oEil9kL
z2PU(_Q(>aB98vsTVukJ5X<6rjkNzqPf5n=aCy1aLHQ9O_aZIDEb$-PGxM8%OH~%tl
z7ko|j0^13cIHpzoa!mqU7`!8JjL!asxc%7t<s=n5LHWoy*&2)!A`8xD#-L$)`z#u3
zy(Sc%@@3DB&*EWMMqA^lOIrBp$Sv=PiPh}xgaj6Sv!?hCO5NyFf(hasc7q5@zH0cB
zcAW!vbKC_9dM6_I7IlNV2F#n?b!UnT5h&jo)}Ok7xXgJ}wiw*)z}63mKk+KJq{Aiu
z(Nv&*fvVdw;5UBVpnGC|4DgBMXc*R}F;rq>U!a|R(+n*eh%~looDyiXevP<vn&cuD
zMGhtyC`kZU&L9p5n@9<uvp5fuXpXxY4Br&+GLcB;4Kk7d%8H&shQ&wo?KojDO_oAK
z0zzJtGZ^sqr*U$V93k`wQ$pcgtyR^}cg`!kL&oRn!DJ6Np7H72@W4t#=my#Js^~1G
zO0Xn35UZ6BCVT#aye4y^j!-dYoO!BwRkr->Bngx<Z5?*uJYZprksD3Pq<xx(myQci
zJW)%AsyGBxi#9M&V|K0KSfwDI7x-C?RkN$U8II`%u~qjwQFkV*<h3|F#gaB^n^G1A
z4DuiV$%xWA^x6)gvWXT0aVMjGE?Li$s?TBw;DOgdEr`Q>SB?T_ww0>>Xe$3NifieI
zW~blmPKCwuIQs#fc|S}QC6eM;NxT5s7y*OBY@>CoF50PY|Dh7F))^1o1VbSIkqr(V
z%uWF)$l{>4C`f+ZR@vMn6M@4rL8!cvMr##fws{~k7j8jn@j&ST+9g$;0m3diOfz`_
z+fBZYOlj{u8^ZTxGXp)D*{$HbvWJu^%OphG*bj8oBh!ZBV&)YZ93ezIe%klRodB0>
zDJ5^U3GPbbZwo5aF|Ic(Z?D9yKb<7$w=vHh+6;7ts74c~=$t2U;|kcr@-w;~b6y!5
z4c%(}d+V^#+?Z=hU397e!DJ}Vc6JZWoqIm^e!alW6S-X=wQ}Sx8+q_<Flm{ViJk6F
zwQl}E6cNy{&t(_>KQ$g4jQ=l<2O}#R)BjcD!N|tM!218G@vw0>vag+cHObCiAI1LP
z7LVNh;Z_I3|Mhk~ed}|(&B?5&sP(ioy18soyMy0>rz}YC$%%?A3!)^ZCB_9t=hu*y
zNKJ?c0PN=<gOQP>x;C~vH?*)LH#|3-1eIrJ1*FL00>H=s_~XYYkN_Nfh0B#qJw>Af
zfHIo&0|m5X3nfc!O_c>;^}_aImokC|h|9{#^3Q6b&Pk%i-{BLPEG-2<hgR~>D2buu
zSJxF#Qk2EvC!{NbLSS!Uc42A;q-e+5P{#r`riqcc$(Er6G<}r?$nxU`F!JI?p87jD
zmVToLa<wg$rTar*c4}+`9gnPph@6<J01ytjPzeSMbtMD9jPRSh(a!Gp70ANk-uRVs
zV*yC}(_{<%Yl;4~`eDC1ez1#z%S=GjH?+70VqjuzYYr;?!`~sfnZ5?J{)5YE=lUjp
zfhRokdk3WWU5H20H?(|(-&`3Lot<Ar;U~1VyV}1ryab7FVP|w{1g6;7!rJ@+H!-)e
zIQsY-`T1|06?MSJ1+kg&SA;XvCzLvCn2J|Q29-+pCvIbE{uC$t7g?3{`=xCxBSWKO
z^QU9?8He|mT^`Ys6La>r`<Tb}+d|mW8Jymp2=w9yzOQGYz-{D(EXD2Q-#EpmiN%%i
zpChWSIk4cQuk4J^_?KA*VD#^33agVt!xNwx^ZeJ02jkc5XKe9@kAjfU3*Jss6iyX@
zmKhn?Kj9<~o*X*q?oVWGePv=|dG<sv@Na3#@7oVHv7#cZg7JB?v4_zuT*cC9U5{^J
zaKq95a>D{x<<dlSKc(Up+lFLH02=vci2B+K^_s^`y*q=Scz0(CcjAkTXh>aY{U~CB
zFZiHyYQ1h`_~g`gMe8o|!_MWvb4}k;y>--4pB>SS?&8Wy07^NDCWG^AMKQLdTl_HE
z4`y=ns~@rLNEIFLW|-1tB7wQN*Ffg2q6)!`RuH;2=3Vb?mEZRa^lorB-ta;T<r~(S
z<GbdgUvgoxqiJ2#i`v6(>-HDUT=lqQo);?aYe>f=#3_z{yzC(E@R;t+-%NE**JNMZ
zYW@SSnvVGy%YycF)N)0FdxH*>`k2I4;yR&_6K~e_??o*&C_7)DXVT|HjjA2+Cxjrm
zH$zH_b6fsheZ*LeClhdBc`R8CX3h+y*C@I(VREU)7xaRRQZDI$6gyoi(EcS+OL$uS
z6f9KWrv1v9_3vRW2RLv-$=of_!z7J2lb1Mw*<&ylc;O#=!1^1w@T7vD)N`%-ZJntI
zfYp{h2S$rRX5z9yON~sBGms>DF7Is>PSg(P?}w$e3EIrW5a&3lRVPsBs@<9}5yQ)T
z%nYA)?Y5aLTYTLm(r2Djc7fMzuCJMPeBIo_)Qh*^11=YgXxj7d(-M)IAg&Rtmdb?O
zRXWTDMGWuk3KJ#fo=}23H#(Y$gYS4ja%uz9YjXCeefMheKx%|#uqq@k=Q7;R1-#J4
zmZQ&b+o}~p$Rzl-v39<>-mMaUGKZehS?V?w;7!UPs(KrUw;2J?_MmyE6tzXP3$jfC
zWVK9c5AyIF^z!Gsiwgj};?;eZgs8D!oOr_w%7?ZeuNuPb@OZ99S)a@ch7a-HpsG;2
zq4~wLU)3g1kG5woXK`-{$G22sjb$?5_Q7mo8lCxC9HOp%7tt%8+Ye(4kAC?xLU|38
zgAviNf%kHh>!@Y}Ow=aF%$2@=840A!o2P;f`%m;(L3C_rpx|f?TVIWO31P0DFSFgL
zS%XpuQ_~1H_D8%y(II3~;<s_H0e472JEa$T@bKn=PG$qy%gv2ts;An<ai&8k#u1tx
zs5GYrDOPk*WE74--5TjJ;1uZd@u_b7TOAeg_66J_kVBAOfM^D+oYVJ9jYfw9nceDO
zpx9&m(_oS{p$X4&^ZHrCopi8{b0;~sjTorcMpkB*b$niryg{O{Q?foiY=q906#0AB
zzh*Os*gOS(zg(>&U8s+fJcWI>Aqcr6Xkf7|EiFM0aQs<uyBGT{s}ZHHGE62t>ctIA
z`z`t&0_CRByWD?Tkh%^36>dy4$!WxV?(`?%+Dm}w7Q|DZ8)LH=n;v~h<#ADSWXyRA
z++&oE`$p!(BMumJRUYPzT~Zox)L5v}PAV*+7D29ZoBrHM4&rztA;ovtf)|5Cld<X1
zt{zPs<NkcaHT3jWf6SS>*i#Q|lAqtXj;_6b4@n{IQZILc-)&cl$4)|SY}kzckSJx|
zPwFZfPG|tI+bSIJIgSqrP?<C1`wc=5DMQW_>`}FVM$6J0D9jmZvlr-WB<n9~L1agE
z83@(goe*2tFt7L+BD-7-XRd%W{*q+0)5NO}J-Vo_lQ5ia>317pTcYC2MCRQhXG~dw
z2{jb~-S&L-56@kdiO{Q1KT=0%FDM=*AU|7vxanpNB8&VXR>XPInibr0FuR|xTfWfZ
zRh?Btp@l@rv=gZ?TC{jJ_3iJP-Uso|&)GNmj*^^38j$IPS!!(Sjy~5G-7Au~D}O&C
zs`d=c4rNNKwVzL8M7M(}Vz>~Wg*DZmJ$5kW?4oCsx-mDpBwIWWJeM_!zU=6P?_#KV
z;y+?E!#6M}28*oKNz|iiUbdrtHJjH_C=a~Sxg@w7xv5v!Heq6`yW|D5=T!5Dr3QKA
zV>+c4!BqV1IK0bf=w@bx+3=Qd!yV^C6O0N*QCOP+>sA!UE5anqYeW~~bI_qV5lk}{
z1?99RF>snI>DdG?X{4JR`^MDLB~8t(36GKk?OW%a)D2G~Xy7qezBUz@<k55^vCS~v
zm8cv0N<u_ckwB7U)q4Q7<~#n&KtLaCJkn`Pd|cF6fVN-IStovQ0vEAtWQ0k;u)+z?
zBx3bLThVFi=irIoR3eLiX7wmWg3u9hC~=P?R2TjAH*?r;6YDVp_!KD61(3dr_|UQd
zBIog~BxI9&V>xS%iJBe|S7o_9<q-T%w&;^K&*ybGhW1a!;hlLe)@Nm_Q-C)GT?zra
zO5sqIzSlU_EQcdVXG$Jf#zN=41<y`o?4TSA^SNonY5G|C^8Y1;exdIHF>l(^d<*;?
zOzRHUYLcxFYjf^=TIu<K<ytVhHmU#K0*e~f1-1af%j>{t!wzhaDVd((0_M>>&hInb
zT`0angBz*t$)ZzLXTpQx>mB=pp+md51%7#|E}%}j$2nh|FCwWUatGzqL`;QzH@W5Y
zXPrjXmII-dmOLQ|+f=J37De+5>zxT|i6$Ek(L-TU)T6;;KZ-t&%NM=I=I^N_pne?+
zIVB5x$zNODchSpMxHZEWTj$+=qM!^^#b-KcTfwj<;sFdaUM7&yv%GL^1_08HelS=d
zD$2$v(qmbbXh+>7)`>{nN_MaF)J$zKtUm|>GAVJI-fYRAvgs7;K@&i{VAm*dIJ*r$
zhnT<P-dUSREUB&^r*;{{d4zIKRQ40<a+Du5{FZASQ7RuS=?8<&Tf-#nyE_9KC@qre
zn$?Vi-m68<ZMksz9+~loCchwz4gNu79pA8jbkOsUnYE~=Li*oN)dg&DilKABFb%ON
zYpb6_>_DaLl$H=<WAv7wgx+^<&|mAE7dcwB1l3)Ep{M_9_4OpBrf1nbm^i=Q{7D(&
zT!URMlD^%Eu6`T4Z#HFyVGln|rfa%qt{3!=?LID`#?ZDLEmVvHnF&Vy$)>Gi9bk7O
z%;E45`D~)Grsx{LM1PM>5%6w=2WM}?hRJ}pI7TiJ*wZ+^T~XNJk*+uMm{swJ5HXPS
z$bH8n3)Ea_KL%QBsg+hLY$?v$U~oEs)MClFjTVsP7EVfh*y#8y;6Q(sMs)(kq8QU#
zti2by)G9+<v|)wQQI!(CGa`QM*|My8lds;bEbf8u06%@3k<Oy<`_#AcF@T?U*2RcJ
zaUDtm_ie4R<R}bkPw}r~=8VUL@@Qik)uX0<`uT{!<p-*iHL<4B8&RqVX80yNb%(FU
z=@ZxfATG|{ljbRkRYB6=7!G`!$6#X-7FsS=1<;j?qC3>EiM8LHR>xAt^yQ!+e=@xA
z#+eszUOlmqowJ!`T4TbMMVwl|BX9kjtUgo=vNMVsv**PsWcOxOcxN!R@&zpp*D9>X
ztcbFhnbgu1d_kY%9q60ry8}(JyX%A4T4@H|silUdtEB&6Ch|3~P0iIA#81^E{&)U`
zKdQOZZ=md2#CvQ9CMgixwJ@p$o}L@S{ryzvEAl@fzGauyM2r}rvH-GU%^aHShN@<o
zGApPVx7TVgxt=UI9y<|S4}G2`D9Q-Ym{9wRuvmF};T|lhloLubX$SDMI1~5`r3;{x
zY%w5`oVj?!dpowOIpG^+Hu2GltY3Kx_tbCVEapp7r>9L;>|n*02+m)!#Csc%9l%Qx
zZ3=V8(DcHs^(a0N6IPBtw?_%W369X&@<_tmt5M3=#|vez3t^<x<@JL2IyToQdQ(?(
z$@-}p1w|aX8PrHxd%S^`>;82@9xRrCQTEF+5Y3ZmcDx$;OIJl+fXShDJZ<0g%WpZu
zne~hlP74LSa_vVcy~ulk1z$GbfnlQQ)Sw9JK%Wy&qoWYJm=P)y+Z}(@sC~bI!Q{ON
zs7z({{36fi&yAw%^zxP5b^<Yp>h|4&_orUMCj({VMcDPpC@D2oa*aF6iU}EDlUbyb
zy7Tu*9@`H!?u!QeniG7*_Y)^2g`xuf7lRm&*1jupRhqlAeTe2;?B$V=(%J4hL?h5$
z?ZA2LDl5Y<3nmfx9ym05LPCK=*aWTMt5`BAfg6d%JfKngr3GuHIrE=%rM#U-t${7v
zwQ%X376Kx_1tnQG?Qw?gjZe0^$hk2wr8K`(8Bn{C6%Fw7;K$iaSHxZp%<m4}O?y%|
zpG<l`KcumF-e~KY(di{rW<MGWOh!nDFbdAF;49>zc1(1&Y`eO;jsQas-wt@u*BX-}
zv4!|8@25`|0@Y0DczM1@zMxV5FShj=n|$0Xmi(h;y*2ij6PuBc1wV?(`pixqAhZ1f
z_UKW@mMAz=a`lt!iZCfCj1Ct_1UDke>LrnCmd(6}$aarfC2%@Dp;YDuDq%-E$vfe%
z_S_=I?<gL?iF*_(cV9EXLV>!Fa*6{gKnTq68&2I6&8<9LU15vPFz}i+SN!r8RL9G}
zc<{eCkFM8Qv%gX9?^rnlSTwK;#wWpRW~A0P&6?dIm@g@)9Z~`uZeT{c1oZ7854@B?
zqe@mgWl(a_i3rqSSn=7KYB1!KHB&Z>ovDdkBG^k=76{`lbzSPl^iEfOd~*B-YiX*I
zVA3vyuidq3CA`l`JVjXOSS20RLcwULL-Xo=3bj|WxpEmBDBf~vw7Q;x!<@d0t{<R!
z{4;6>Wf!N<9(Ma8+ZtI$bs&eGco6<W64>X+Vln;Os5lr$iQ}JC`HOn*I+5_OkJdXe
z(JA0;Et}E`>u^-{-f-`A|1jh}3Xd5CnJ}XdK8ietGi%#E-&WVNCSuJOm4)In<OGKa
zdwk`@4a4eSZ)|WaAav{y?)U2`o2R;Aya{RFF3$Tvei&XXhWM_y+L?Pt=Zqd4Nss6z
z!UeqMwMzvnj<cV(m=1V->x2aF<L{?qw!yn_I~tiNFNBRi7)|LtT-MQg__{+o9n>Ha
zE4u?(VocLKjuX)*ErLRf&AH4h(zYaEqkDmyoMRPl7cydOvEei~qj;%aH{o!+YH%!i
zJtqH+GBf9PI5xxlWG{b)0Cc6J_T_oM|D2pK;vt+8^xN<!CdgQV`w%Oi1^&-41HOKX
zpSs=?-@dqhJx!SuOk*R{pS|S7$2kU#eMa1LUJ0)P{K4k;u&RnB@Y@Q`s*FVo{iBN3
z=G;FWw)9*p&fB06Nq$MY$(60$w*AKmDJgN<&H(yBflO@X!p@#yQ}VV5I8bcC=p%5`
zXW^o$|5Puc{RNa(TxTG1MNHR{ELz%7Ps_58nM9gm*+ptb7Y8OZym3Zv(40~T&J%$`
zuw}D>EV!#_uFlvYV%;tV&>ZX&60<1}Qd+XpT2_QK;MOSw1S7xUg^Tj#kIc4bIV_oe
zfAQ9{2~ya`V+Lf7x;Vq~Bdxl8Lmz^0v<x<p%<k<SZZ6e^V{`o;qENl)3ZwpAfPW5*
zvv8|d7mSWJf=_aAk%R58?Q`_PoHt$bR+difdR;!!<Lu~heisi2x)acQv81Lw0!_a}
z-xhuL!<yP)St!Q!tKCB_-MTorG^HboQTR^(UhbdHRn9i=j$-OB93F~<pPG9b9o+vd
zHX`qRnCRAbWiH#HsmRgIgN$reYfC(PoTzBt@qC9C0Up1l9pK%eWmCo$kEk5pLKM{l
z>Q6|g*nQ5KEP~0{WCuHIrb`#G?$ZXo!|-M}Hljg#(0gEL5Ef5ov<1AS>SJo7w!??{
z32qTPGzTUgm84c`_CC&0EuEO=YTZm4g>_6ekHxq{;}Rp3aCB<pDt_`TWPR$FiQpwb
zrDqlU=`6_*VbDQ@=~g#Uwk|sgi!8h_rx=8>{kSw;+N?5Bzqfb^tw2hk5DrZI=Vm%m
z$`-`C=OCiN26Y5lP2vZsc+{Phv?`t|3QjD=g}6!-4x%<l?1KmTcB%MY0k0@R|4(dz
ziYcln*;^&s@?kJ#+V25S?%;OBDxiQ3Mo#311V)!U01)M;0Jw=*3@-^6SJ9Ajh@+d;
zT1BR;CCHMr6qFAEU|Uy#V6*hO04^4Z9>jH--c0%Si*bKmwk*W{_(YnuGr8p~=^ZyS
z=g;W?<wiqcJ(x#p<cuA66jyw0Qom&D%}v-MvP0626M8uV`9uA$VPdsE<;unF*T6@!
zxHV<dRmwG5M{Bm!{ha8}w&`{w5G|7gGn?>dmz{bc?mx~zKM$?HS-NRV&4fH=EIKTC
zOP%IWIEg|}O!r$?(?1<`#@FK=l_X}!n(ROzH^=iBH=t$7-hc`7w7~7*iHk^&;Im`o
zZ^0`}9&@pJL(m2RAB%MuPVAlE>QVW1s26}uJ){pb2HUl}rIW|c%X^~BXhecj8piD?
z3ccOZAsY(Z>?en=Yo6bhV9wHoeYCq5oB}=6*uS*--Xc0|F}~nPNpi+OxbPo92e;HH
z=&=MZaJt%hlRPgo`_fNzW(TZWUM$|1R1+|P@RfP2%uW8e1w6A&j2tfK<YIearZ*jd
zz3CsaffPK0@CE#(^2}T?=-Bcxu9ij>mY$mXenBH>p()7*{ea4h`fqay=ri}^a;hR&
zn0F?+vI{2Lc^|ha>P|8#YFRmWp#Q2B<FL+G3EkcFh<@V`9z=f=7{CE3iwXAjQHUk=
zaOgM{Llpsc;(12IQs%hm1;$@?x8`oU!$81TMTi(eh#&SJYo>(4!DO}ZjV?T{$8Q(9
zeZBF7@pSF<z&=|4P?`x`Llgv#ms|TwDSsXzVF&r{`YRm7#!|Ku(1em3EWL-WK)8Z?
z;*;oTMvl@!%7+_jD5cD|pG;~>XuWU(!4)+R<c3jD--$5~7al_ryghW&n-E7G@IJ^}
z@K2U#=aW#Gb3TqXY&91!zc^Btn4@s*CZS&(Rl!%0L>pqD9q+!c*^pF66niI0>F7pB
zMYMJP-N|W|vuG7F4LoS;&2pBGb4(U_?_#cdM!+r8XQgIdAai1Em{H}hCL0f*snr3-
zXfm|S?<ifKsG#^cCLU`HZ@3oofzrkqd_gm0t>lWtiyOU`ejb^MZG-w!v`7?97GyZ`
zBJx5#v@gm?PW6tX@q|eCQE)ZHojBw#Z*#tverG2G>y7;0`2oDR1Hqy}JjuF#`WZj9
z3)0hDggx2Wnf6+7?2XbP-?nDvopt`^oxRkt!j*zX`BJPih0|EA+f+{qn;rlb|JQRo
z_Iw(RWrYs)RtC++Yf#+=D^k5v?K36?#7f0!91;?Ms9@Y^+1vE%aL8te-%s-macQEg
ze?jU;B(4>`5|Y;52&V3T6UOsH1UgWT+Hd<P=(U+{=)IqV7$z#YOOSpwZcOA)itA;>
zWMv*IR<vXgi4Nb#KTm4c>!-yZvZUN`M_lL%N=;1A&nZ7wk8-*SO0GzJHYcR#<UBYh
z-|%{4Zh?9+d3Hqt!>)Q#w|vqfIKV^0{tIxGQu7`8E5!N|{Zq|H;wg+Z5KYM;NfX*N
z%t|N+P`_MsynFm+JyV{3SgiN9&ym?~83E*)y5w#bWj^eTPKVR*H5mKX;%uve1v!}P
ze*Lb!>e*DwI{4S0$T-SFK4w<|o<Le`Ji9+4j>Cj01ewu`C+_1o=tPs2U9OeS2mZ?Q
z*!9YPasq|aB;0*Ve-3G|wg?B-;T+5JW#S2#XKEDalHD-}h$LnP3sx!wM9W;czcl?X
zdUrOdr>$N-Z-04xB)C62r?@o_vR7|Fk;A_vQvIK%^}~WL965>Y##7>%vhsy(gKXFx
zb^B^kD9ggD73A}fBq@s&QZ!Vlz6U!+0>gQVnCI1O`_BUr`_-D)v%A*$r$<EbvVwG1
zS8uLVoU}+;kcyzJnf6@-{&y7?P!~&q*=!_}3r(Wy;v7tw^W<vTrsrtNH<BFvPI`zT
z@@SdkAmyKI>$shlD}TkK!UqbF(xMv&)5nfauBfV2zH1Iu9q01!5=kRD^Ea2XeUJ<i
zNOUi!eVF`G`qhUCYe$Pimja*e1@qFT+goOgNiN{A=6nQWbQ7=3W_Q%#oo+=cYzcA>
zGJc=s=`ZVib--lIKKwT)UclNhe=$_?rYF8+rh4d9He0IUvXgp2RfiyEBTCMrB$iAP
zUp@I9GS6v!du_^rRTbr+x9IS?(^16uR&7N9rSP_ny5!~TjE`ZM@G?4&d>FaNy+9eY
zNg+`{E&Zp-Sf3pUfmEsB`l7uW=in5yw~ibkW1*6J6qRTueE;iXTGpZLFh6+p!@_^u
z7^410;9(A>Kn-a<K#l%I=osI;!)@M7TSjj>Bs)%UhKtqnM07>H#3TCQiP+`O0uxpD
zGcRqGXkA5QY|fDpq{c;jiJuDmPFTWo3)xu}$qbGP*GR+QD%~%MG|#Y{su<BM61Mc8
zcDH4l4Dbiiu!<|5d1jdImPw<^XZ3qf9Stspp1SmqsFlg22$@87ZzS<?aOK%{w$?Au
zZ2@RR=VcuZlRSYPiaw9aO5^f$a$XL`5Q6W))rAJvCZxt6gC>UP^#~WUMuWD8!!5`Q
zl{sa(i6qdk&qFFkZc`FPg6~6-tgyfaieM=P80g>YePcxsdcR!3w=hmVUS4>!>Uu99
z!<{r9ioMe$UvZr8@(ST>up82@C64*=I%K<Y>@b?jl-Sgcmq&dPB6{C0n6d`6|GqcV
z8*=|j4K;xQv6?FDUmDdElgT!Efk-DuI*$6zX|U+6mPkkT!QETVLu<w9Ih|V~Xe8>l
zgVj)mrQ3Q2RFsSlfDF6U$7pHG?J?7>Fu4TRvuRXn1IM6kkkpAFDC+|EFivpN{PY9L
zWw6p*&4DLR`*QNgNPWJmP3K-OyG#L^AlV2b;ffUHOH~{)=D1oQpB(kdLEhxu4*H$D
z<;Boo5?WQd*^NUVJMy{ng3cv#P2BWr5I2m)rRvd&(K^pIk4$L*`81=KfWFk|-$)5R
z$YB}S0#ePl8iJ*RG^zm}Sx=8GN4v9bSb~&2yx9Zgb5~cT+s4lx2&e4}3}h7Fd8&C&
zoZ5Kj!HmHc6?Pv3>#}tTe~rDBG8^*E)2jt}5;Y8mlVi*E{My1~THiY*xvA%IaDAXa
zxF39G*3`y0F7Z(#!a%hBI%<7%tvZFMmPGt1Q~nscG~5w=4<iibX1d@jk7RgI3EAYq
ze4_KPs-iOWi6Y!4jS2C)a0HW@g1qwt<TsAZ|Ko;*$bK;RejG0wmURhP)j-qA0ceCp
zGyJ7U6~jVCI`A0dp^_PV0FGp>2ZTawC(=mIp0h_27Tl_EE>@R5?e<3cAf#{X*(}Ih
z360hXw%@JkCyW9lN{G6=Y^iEy>gvv4eHZ2gzRR0ks$7P1;Dfyv%EP(YeHyS$0<*l{
zjr>DM*mZ(N{pY&ZYNm|e^bp{e3+0v3G-|)67wHhyYDkvKY_~s$VsEQHhAw+7pTYni
zPqk+LOeW_E!#pFwY&$6-ZMr(Cg__=H{U53=dm{1(92{&L|Bi8+-rBm`yu=2oiUHs6
zOf9^2+Kz1=R?@Xv%Cvn>#K1>yTr00bs)~dF3stnIJhEa*sYSMCZ+_gPTG4lLfAD%(
z5|@<kmp_d$+xfUU!GsR(^gD7tcr*?4fM%%2*Zg{zVoS2Oj>}KRE_zS6;mlg4aKxlZ
z&ssD?lB!UgH=i+<g*e((_X~p%SY*!*Nq*rF3b+z0PtczTqKEUAX73hxqD`5OR#B!u
zBj!V~P>-X;97+xzSxvUiON`KrNT?8&9UD$Ybi|#BLeGb?0%gx(l@nmYJCdr9af4Pm
zYpFi(8w1kI3NK5Uh0;W(qPrDD(QR>f8-r`^?TMzG*Z_U|{cF+JPE7pKi<mk0w-YvW
z@Tg8h3$WbbA?!B1TzC`aLU08w$ts^C9fvxh4$cg!aQSQDI4eWcz1f62FoQ?+uHFI#
z%GxI#N)6ZHAmn13+u1%IWli@ZvNi4<u&JcnP8=?`mAwAn7=-+OshP``Ke{lE4b{?5
z51R3r%`cGe{VsQYGwK?7bBmiGr@fw22YYsgnp1ax%pg+FZvYE|(L(KXuYEiDtVg9f
zYhnw!pN45IK+}2&3|p5z`o|)<*PgQS;~Gdb)=ea^2meqaZKQv!Y!VpW{4aiLK@v6(
z_PmgCcXh@<_=G4@7ETwD2}WTI%3654OT!_Wy5JB=cctS`wzzfX&4lcHqBs+%4;W~#
zne+ZBgw8`rO&(R*?2|^aqiXM56N%@%))tC8{RZ75-<-auHQe$y(l`+e@k~M{{tZgy
ziRc$tRYXxAw<9-tISG+6j?kPj?L*WB<^dv@Uc)nNM6#siORU|5&rtsn7j6B`X|{>w
z?Tj3-Fw5(v?Kz!iE6Zxr2vO`Qq0WQr*1aJ&8!dA+OJNsRlgl#xRq+_Qpb+xs$H{BD
zu&x{*x_ydorTS!%M??OYB>y=STu&A7VF{%P-vmr%(^E5=`giYw3$NzCUF1|IG1|JJ
zm-)2|h*s_uXl-d8%vb~9p5c6{JU~z=eP%kbW|xJcyqDE)fs1msjA%;p=c(Y(ku5(4
zrncA=lEjBE>UZh=J~XWVVeA}&gzKVYTefZ6w(YNM+qP}nwr$(CZQIrV<jr10_v{YN
z_(trETq~W6QoWog%Q8_A$Tn^4GR#VM%CPvq5qj`#@uAig@aWkO<UlK^(yLKFd`oS<
zQ(3pRvW-1_7yB_8z=vTb(h@NCZV+8+qVdL`2F_J9>6M2~AP=uLx<i%+*Kv>zxW@(D
z4P|PksMZ0*T0Yhs%+gj#myg@;sq81^yeACUP&rz}{A+Md*fXzu<~khKN`QP^pK5rY
zz@v=5_MH^JzAU_4dFp%l)1#-N)9YcohXD8<M$(oFl{3Ck)V(OdYEW?kgog&^wFNIK
z3gC;cs#bUdipQ&RmZ%uQ_iB#7<*J;mw@u#I5;6(?#Y9NK%T5O>A=iWu?Ir>k!FDz~
z23M5TCE*G(oYSN?v4r-qgIGL6&(j5H$2SN(9gTuUr-{k<e5vMjWP#drb7lbT{XP}+
z5)=U(AKX?GCVq{@^R;St3(<4!m$i5W*q|dVqBn*>DCbB~auJf3Kz|zOg%aC*`owFs
z07!As=8xACPaNAxr{+jkWbC)Xp;14W6}Y5fRVABDNHiq?@mI;=rzW|PXp+&S%)F~b
z>(3+AUVk}(v4~=b|F?g~4)}{w=ViRe*=W8T0i}|(aLt_H`$Oyo6xTqEgSndT1-&uA
zE#7>0sZXShHzjWjj6C(UzI$`5C<TN+Q^bNUxrgOD08C2qE%t5F(F6~NhmyUdmSgNf
z(@0F!zm)QR%Q7?X!j1{XWHi~BG-RYSUFE6h^<n^j-oR@Y&nwlvKm;>#*|q*%wP3vX
zZ0A|pUHLs~KKoC;6xntje!(s56~&IWzCiTnP^eJl-9;PVpT1C?fgkttOp+_`rhUzU
zM|wj9(2+o64K57z9oTpsUOH{Dpn+_TTdWxKW3}_`>y}F{W5nc2O58IyR7spOeUJ_=
z^4gV$-ona1bUK%1c-YOH8>8neWsJAgADRFL4wT$|y(&4qX0b!nn1>*{w!PJF)Luql
zLoYIT?-v4?Tr^=(c6=NQB+gxmNSR_P{UoKY^}^(fbb@b5T8Hf`yMreC1XX>Kjb;Bd
z%OE$zp6;1Ze%Vb4vfF)~Mqp<#7V9S<VQvJ!*eFEuEKVF4lTx`76}gx69Yudwu$i^!
zIw(Dkh_H$!AX}74yYe2WvV4#AT((#fy7b}&J?ZQZ5k*f7+G*qL$KIa4lu>Hs*9m93
zcW?_i)|SR9bp19ga4tiodBf!M)XE3-f(+L{sP}?~5hriiS}k>J@%eZ0_0R>H7%Q1&
zb+Q?%^8-f5qnO3tL3dX!?wJ=fZhZY=-ibH!31S8}@KPV*DvWOLvL9K-fyMv+8fV^5
z!nX4X--C;gRVRv!ti=%7lB~{crlZp1AbNV;tP_ps5XwiRkt&)6ULyG@vFx;fr5`({
z$>VE4c$yiLuPB4)hV&kPlXGh4e5h4EthmB&LvbNYdB(bS^mXZCE#+MPX$dVT3-^WN
z3=AKFJJPc(&WKmF-J+<}rOgc8hQ~uGjvjj=wTCxDqr6ep42;l2F4$pTQP3FWenA~(
z6JP*}<ZQ7pv=LM9XixU$V0B@qxWUjFf55z7AVm^TtV|amB5E+fRdF1s>n&Y{uvY<0
z#TQmQbOyDjH%4s_ISKfQa?xjB>PhB7L=*rfeXnf1KsO@t!%<~BS4x|U?~z^|Tt%ys
zxNbr9y_+353wsy!HCklMa9b0sw6%M}70-CDR=)rPr^YZWOBw7ZvQs+K$g|!^rpM8@
zA6q_fse0N>?<uu?dL%2GnAy8sVPR*I!T8RjW74COySmxNxG`=Ig7$qzwQu)s`r3a}
zs<1lh`#Rz6Br@}?KH-`yRncxph?tq@<s&nOfy}SLg*=P<zlv&nW>$pBsJDj^>SG75
zMD`Bn(ntwor7CKGATle*c*c;B6mBa|l|FgsM*0GbepH{5G07{NcU%KUCWHGHnTY0=
z!y$T`h1C7BgDOU^o+hH_+C%{*`G$z0tKpSsrdp#!0r6Na49+G)i0RnaKh4P^d^4w2
z`j_(AG?St(G)F3<DYo*s4kW6!@U+iZ6Qn!3QTXtt78_eTLW+6KJy4B<B-J|>-J%;u
zUP@bGSpwI^)xg7;xI^MKp~NnTJLbT5l+4V!LL^cDkut=Qvc=#WG$hc1P@&47exv<d
z8FwBgGEYH_zGh>ifR+y&C>A)(Buxk-;^l<dCyu%%+D<PMfkXDLV_S?O5oCba4Itq|
zfH^%wA`f*1K*D&&a|!**ixld)gz_iP&D9`y&9|MJ8Okyj8pcUpJ9G3yuR2N!=p2sy
z3^9h&zMbbMR*U1?1u&<HaKAR~rc5^`j<SQZ$mJ)tSxFqz<*DwWToVPbrsi8R8n4|R
z_;P+X>v*C%+@Tc_m_V*MFSrO%47&B^3RrJ>`;y0icn0vDW}69tc>MG-&0M~;*jRPR
z<2HK_uO@N%mCVXAWR|WUXqcVN_1&W)kEL6MDNHnQ+EOh@tKnqlLVZ<koq%}Ngmf{|
zOwCL~k28MWUDG@W<VnaK5XzjyqT1qfx+?y1#N+<>bvo1M`MKRPOk=p~1evyXf-oeO
zX_Yl!<>n~KuK5A@O)&Gr)r!#jW#PW>&lrE2Go?Je<hzH$MhHw&rl7hyHx*S5Yay%g
z(Lo~b<D1Q}(hR+gHlIgg>&jgj%;Bj8LTg6-T>gZI2uv+nuk#wYLifG`Qkm%Zz`YD`
zy7aHic4ASSC&$h&Mr$DtoXFP;#jP;AWZ2VNFe37o<FAB5c~_4U*pkl8;lf23XBsGi
z9EKf(Y;M%lJHechr^erTU^6y&A|)|k{d(b&?&>`oLPs4tH;4t(yUMw#{i3voSyqX)
z!pFE>kKC#uWcPYaVUFA37a|;APV!ecER;_GV;Y?|DCAE32jo^o@>wWKN<1N<NPxy3
zLpZzRF+g|3GJD$UU((*Ew_Rrqq(iV#Gx$Jk`HxICW3}epfP$O*t@BvxWceoPi5v9M
zchXw;-;d2{*O!Q)@_KCLoY3rVv8L_(uZbPR;}hpOPlx>OsAfFfSZz4H`_l0%|LnG#
zOi}-f)JXPDEcg3b3?pjSYB`;849iGqyj7x^4w-H;ctDf;=;Fc;MGlY5u1VTuZL<cz
z?H#&hOV>9WKqpRHz?kzc<>bl_wabho7(e4!WyZj`Lm&vXU4Ry+tPq4Bl?O>38gC?@
zgg#nu`pU%FD!dd8kT+CEs{I1a9z*RnsUWtTI#e_Sa_IC!6Pjp3ol&9jEVS7?HD$z}
zatI0rXBVBu20h}NHctJ!zy}ZLXxCw`nc8@L<SU$Y0ta+0<Q|w@r#Mf})P_qs3uB#5
z>n%H?E`5c2YRb4Qd%NiC58*+rS!;t0tQ?MpQSSpD)c)I~Kt6{Au&V=JElrGShIK(l
zIXmCIG};*<`v4b}tZJovKWu_*{~{#t7Uo|s!#R&I-cuF)QIC=P>|@se!R`6r#B<L_
zQ$@$7syXFys*I)!IvW<YO`Pj0D;2?!z&;tKu?K+!uUg%ICu3J$TxX~n%8!xnvCVX6
zXAH!oYL7|5imsaUtr2sglE0yQZ?vP5PbEKBLCa;wut=c&E#1t9n$JOEEj)PLrkKc|
zZT^=)NO_%V`|NZ^sfaIz`#!V#>Ewi$X1WT;`Lj<Qe7h6`U96&5#;|yu+j|QXXP4s#
z@{%Q;hk)Jk`%)-b+dgATOm<jV5|rb53_FtgMzaPHK4ZAdnl_(O01dB$8j>04ERDAw
zQA6W-v^Ko?>5F|O)njl_$9bladOhuQ+Rv{@S@xw_>bu64GCx%!U$Ep`qP?~ido=kg
zyXh|W3b(t0$Ai0Pv=O>z6%}Dn8FXkXNjoBT(imx-wHf5OF=r!n=4J5%0>G7Yen?l%
zXq4W?fJ_L##v&}3D`sEIZ-eJ+{UX}P)n{>-=r(<`u%gPGbf?HMUdbdXqq*x<j$~iN
zDWsiWqyfFgl>WZ=Q-*VpQ@Kkyj@bVG4>ZZM$x*8RLyI6ss##snhKcjOVA@(wYvq+m
z338wBZIAo4>t%y;3`PVoa2KMvoI3utliYYE0pankX@XP>p`z$?z)h@`qk28h>!Ys1
zqE(*hN^<13h3}{bVV!LDIMfc>_JAUiT@=H|ho&%$)RIiUJRzQ7+A5<F-m1P7SY#Uh
zC1~1!AH2w~J`~*4M7t0cBqPA@#d#tA!8S+?PVE%#{sV?*-Qrz73g^1nRXl&l_4F(x
zk?Z<zJe57fB%u@`4A{gDP}as=na}zTc?n#C+>2b{D<geWAL~+XG-a1$kK#)jili3b
zJ04a2`j<tzYhqgK45e`D5S=7dwr2#73Q3Wcb7uYSRk{*oP-P`*A(i+rGRha^Xe*UF
z!hWY?YSvCL!N!OSE-`)VB5y`+nr7}wgh?SVsRiXEuQUnySmAtlftIbqc-0RsI>9}1
zdff2c&Qyy>fRE0Kh&hUK&vhF)YN6({c5sow%BIrN%oX1`K?5{618}DT(7p6u2{k=m
zoD7N>i0N$Rx&ugw)3a<Vu<QPGHTdGzZmjmt(i4P#JI_x5F@II?OVW-~P$xaugWtb%
z&9j?5&bFT6WMpwQ0``)%jZ`TE$mA4hUiJsK(7|WG%f@@UR;MJCeYj)-4s8)*oMC!?
zrF1TP_U38DyNbYdStsoVC$0~^{9fSV$&RI(h&{Y@*|>H2;OoPR3A&Sl1_Ra3mne!9
zsCI2kAN-*mHty9xw4#LDdjTdxe9|bva46h{#)3r_hgg_Qjl?<et88T?kqod^c0M)X
z8cp+NIFQX(7Jbq&PY*-nXQQBm>F-xCYx=Nst%*1#w93ULLgFfTWfLAxnL3Pc&rl+^
z{W+;%ER4vf(LcTtFX<^i>$d?nV?Md#K--!dQoIQ~AzS3*P=RP+jQE_vr!EG<P*Du?
zYsRiZsL%ro`Z3jZd6KGK;c~%ODYzNQ2X8AP#!JvOYl*Dd049RfRC?Bp<ZTFIoxL)U
z1O*|&I=V~SaUiw#78V%ot?tUoq9V>({BKhgYb#LA%}0K73-qb&!O~c$0DpbJ#PGG)
ztHpRkUPx6{XNd@`d(@A6XR4sEOjMlGh<QL^#u?6?Q*5a*E+q?*ERLSVS_@FpSDYAU
zt}(6tHoTDZuXgw%(*td9!_En=KrzVh%oILqx;fk!xaHmVpCRsVB_Ud(cx<R2?%NA5
zMna+Kzm+o}EBXHQdltAlv1FcKvG9|EC9qhW#Y(nOsAdJy2^UzaASOoJ|LC}P=tQ)!
z<FtlzbB((7H6H&AK-(;$h}bhIIQjhqogo2@P=?VzYb<gzG6Gu>qw)n4zL~Z!Y4fqH
zF~S_E2V`F7oZ(jke#?;ZRSj+)QPQ_46sKM1So}hC)*LF|&R>OtC4#XzC}624N6KIa
z;_0Fu`}4RUAaoO`0%xbaOY4cIXiE{#`pau9{J99{cXAo>u*(#Iu||kH9f@I+2VAv5
zGqr_Fl_DHQ!NvyFz`a=~WtGezD$*XJVXZ$Kh5FX?uV+hqnW_-Pi(CoHvTI!9y=pwi
z1tO5feV?jyExhpoY9pLtAYH*nGDO?0w<(6B3PFAE%Mq&i17q6dLa7G&98MjXhqpE=
z&Hg4ZDzZ^5r^Ce~j3cjPPVVpTA#p&DFRvKHW@wzv{7S%z>Yq=$sa7Snx~~^u(v4qB
zNKY#)oi3xOa%(4h2NcelI-B=LrB2{ml{av77n!3c72E39U75H>c>g2ozRUl7Y-fP;
z8srC!>_gxO#)@ahCLOck@l?VJ;~sKyFT1XsS=OVx%25LJZGc8i=4OnsIHHoZNhefT
zK+*%3b?X_)1xhJE1G}mstw$F)x)hg=rDGTU^pHDvY3ZW>AV?(OXtig0n__fIqW#mU
z(dLe+c}}yRsBDk*xnPu#QZ~#<Vz%<9yKwk@i{2{pz7A!kkVT&txg}mbo;HT`roupq
zbeb0Hf&)fr#5j6T?$)$;`kxw6ahc9%il*qJ6s#av%*C%+jwx4eunBlm$ruSN&sWGm
zyS&W;;74D`;vZuoCIbatr*(O~b?8PaP;d*;YbkH3Fv=cIV+R99AhY~nq<dF|LdOmA
z(I(zhn;Fs**V%|N8R@WQ(35y>U(|P8$DS*cryu22rJRj%8Irn^-cXOn7dD3Eh@2V~
zRawciCbH-jhpHNPqKSEZ)I1>ZK#YsCe&Q7w_7G=L2qsYVgOtm>t@FKJLbcXQD5EhH
z2-9+X$8#h)^0v<4nC<Vtp2Qz<SKe@Cjw<DZu2wkOFwpe<a28e>hIi4E-ju1yTOQF8
zXZf~=qDm<6qp^uueHadvF)W;fk15vf#4;4X7(SEj_^8}hw=8mk4AXS@9Jy~Xw8?j>
z9gR%4#WL#K;<O!3M_p7___^p91{UiMY+;kQJSP3|iZF4aEyjP~zVwK9VN6KK2!)do
zbQihn)xmEtOG!1}EMSCVr(T}AAKXlgvw|;tX$Gy0n&NhMhmnC9H$%fgv#b_JPizUQ
z%*_81>xZV5q~i2zZ~s}`#?K#W_TsrBvr?$LkRiC2y4!K8AmvYzMM{!|@?y^D*bGNL
z;cZB}CVS|s3SqG))l}JOEtyMeEO&Z{@DhHnTY4;Ql*v-7El`wt<*h*EAbDFV!RYMn
z^5-;c4fESnh=Veja(G=XxN`6(9jVD61pNA}v(-m4DZhTI@EJOWIZyErK{95xGC=K@
zF7HYZ8DJOV;7Ibt-mUcdvfpC$jOxrAlw*!$?*x?zx$|D0CEWiy&1v!0OeN{l;?bx*
zyrd>P-dKSgL)({YQEu6k1Vs-W8T%s-F(*hjqNtZv*+%yrl0(qgIxkryd09;74q-x`
z#b1-1k7ZpY8KN2QA^ms?mCvE-UF6{k%l26%Xd@GUA*TGJKDg~0zRT3gBQYFpkP*vf
z3zwr%mSpYZpD3&abm8O1`6BJ{V9=C=i;{*+TV;xG>^O0DR?ljHoTh~DezBGm@eLy~
zufa)MVf@JaZ(spX{3D%0<mJQ&5FuM|ZSfuH!g5PBZVpG;TNP=$s-=^qEytl}+#f}N
z>9x8g8ast^%v?$EZ{*dRg*kI2XUSFa?Q=b3)b=Zaf7ZVmpl}T*9`x}cEv3rJH&*q}
z{7s_98tRsNfGGR$c0mOUG$o;5;h^j$kUyJ_f}5wtW+<eUB}ilAlIe_V+4PV!O)QSS
zD~R{3?c=#BNufL;Od2&W;=;37r2`6wk??xJIjhZtlN?VcijOz~A_NMRs|YU%&1H;Z
zcm;A4K7!S2KpR-4id$3)uq73%`0Rk$k&HATSm?j`>3ji5-Q0P?je7V<sPJOs2R_h*
zUnn@FV?_Q}a6#oPzV_)F^j^z<{vN7w#qR?~SX7lvM3;^(7f3P?BQ;P(CsQrIC&o@C
z4`j<8e*Mb=<evheHzq#42Q~uNhRsqz3dVTe-I<OtGiX)TeP4q4VV23o)7Phun<-7Q
z-Qddf<_NN~8f_V#Oc3n8<S5bb{_Bz{TEzn-qITzHwU`ueZKT|hXChBUs}lT~%B45W
z|A|Cb2ju)rbE{<x*qQhTYAHQCjB|(VgDT#%G0`=#S=qJ#l~rT+b3w<<|6%A}rd)tU
zsJC(I$Bsf0x+R9ElsSRVS)3&kg;SP2G;cq@0oF#h{e%Vth~q;d%r0UNXnO$u`?;Ki
z5n<f)*s?qNGKpa8nYHQX0KOaAQ7mC2&^cS!`*5%?ViV0P1qb=@xvF|zZ4o9mVi1H~
zH4*x$JFailUbz;+>fuo*;qf}uZ%I3)2g0{t%zDNm`!A0gp_r2zuy;~vsx>tN$$m#<
z+~rCbZG~py#}!shkxniZJFnP>4}bsQ;grr&S%iqbO#V^pAKK-T#TMP5l7}$89s2t6
znUPnr2#P=%ji#-=D86oTbhwJ!f%&uy1Wx(mTnbca(5-Kjb=i!lpp+D>oRY|0FXWrG
z@ip#IJcsy;T8f)FejRCeo7=>jA`<E;Wshp_5j*D{$oz+oN+IUI94PkT?NpmMI8ca1
z*qcd<RGoQH3Hz<ee56Xp(jFvK<SC>O5B3W_#d0L&tgA7zjyF;!a=Gf=NE0&vzj%=j
z{qMh9QoKFhFfi}*0_9?p+8}7VaJjMfn<m`J-oWTKlxGi}0OJbEU`N4#@~!5R**Sc?
zAA<plHz-=|7xRi{8-Auuy(lH7Wq<}c6=a~05N60_4c)O?DrD<Yp$~_V-)}JrBwjXy
zpfQTO0=c-Q)qv6*wGQ)uN#>^871W2LrEMh>mDO0SC9A>S5Y1txJ{dmTFv0Eko<=Mg
zs&%utzLps4O1w7@{YR8#kIX>TG<oRiY%SKQRUPOw_pjuEcW3#w3ZpIh8p)SJ6FI~N
zcIJgUnb6=eXTNNJnP^Y*zxGE`n^0=IMW{hUe+%u=dIPOC-iY8&?ZkhlHZi}gz9PD{
z@WbRe{0h%tfLh!tQ4=MPa!P6n=&)~Ulhef}4y6W$wN|#=Siqow7oeT%Bo>sCXRGrY
z`WwX&k5s@%{=P2gV+(;g76d8aflfJRUlJY1L=-D4v=`Uf320WW^RY6Ifg5JcJaqG(
zl{m~>{-eI3-0nnp1~aCnY98g^O|h0-en@X^wf;~Nis7wZYuCEsv0yJha0T7h685;<
za#}^Swq-F?Gmo=>Ur|QXky;kvh!0X+Fn(o<`tm=)vl$VVHRZu=jLmY?5F1EdG^k$C
z%rSvEaCqJO@!^86KFBT8|E>mZWWlZ4c>@=v3PHsVMJN8Dt(FgvVKyXzos}lqpm;U~
za@RZgfdTtccQ2U!`8B(-f#zzxV1^J_Re5|eNz32mnZX@Hh2d1*!9^|+%*P~^yU8Nk
z*!)MgLX3#4G`4mr^RqN9<q7(g2}IhgXh_I&Hb(C5JD!DI!ZjIM!(hv@tFc<@cX6^X
zxi5C}RvdJ~6$T;~bE-8T+||c%>#gu6ZCo&rbHqvZms0qFrS%dvo(5DBPto=wfgIKA
z*B$Oe9|+MbhrOJy-ppNl$QammVan337hEXG1wABwBm5U-74Z6p&KZMN3OQI>@q#=r
zk_4l#uurnM12N)i-I;3&3^GlLL(pP%uZ@ejLmR=UXQ>Auo>h`0snKu|q-UY$@&*vg
z(fi6B*lb^$sqg-<>67Nb7bd(Mt&ZA;my&zdgJR`%A1gsR4Tl%Q@#dZ&OR~tUGWL>-
zsp^6aP?8E7xh^RS@0|gD0wwsN0#J)<RvOrj&kXOh-)b)DD1Zuu;Z8+Vdv{BasZA>r
zmsK4%<OnE6P3KcjIL!ssNAS`~5ym%vy)!_`jCbYFYBLXLd#96ow#aEiW{M$pfd`F0
z$_$u)59=kdc+Wz}yt+@Kt3Q|gngu5SYFijC<ugKy`8X7DJmIbocGABgz{v_MtCdOw
zgy@j~%eVp$KI%yv7$1BDU?CeIo5ceyD#F)VICmIzrR_tPqniX2SCYgbWED$1vSfkp
zZz8f=*7N-SkmWa{@`l@gyZYd>XgE_#^!h4gWo0r&c7+12B;{>;taNkwK`Gag<apqX
z(I_4!yH3=X0VDQ@p22|D#C{O?NZVdCsWW`krMEvUGZ#g-d}an{6T5k{OL7(b53Aa|
z3W#?&4TH^t=O1Nbbl1nUPF#?V#eduh;VwH3wb;%zVbpT-dxqCv1cy0|-J;+R$3?^e
z#e;OZU<14zJ)qig`6=H4LQKy=i4TzM8+#FJ-I~1-B%!-b+azg_j`m;&0jla|i;J!#
zmxM_JMU_{+8xEDeFz`sPQLIT?fZ0~jR;~fECclB7aiuv$!kNw*Au?i%^y-HAV?{lC
zwG_`7!OsmG@kx&d^Xd>;OyF3eL(txQe)YO5$aULp1ym{2O~fJz#=&e7vTfd~TovO5
z!x0m<;h-CP`g7F7lasm7YMUaHb*1A9orN#a$d}vY3hmyCbv^^(*2%LJh3XaHc0uuA
zQ43?8?$GtZ)SE`VL-92ki#CSzw>W{p>GaXA7z5^eDwrdrrEMz)v-zK|MekaGMHspB
zo1qmg$37RiHMMq*q>Z=`>b5h$@$h^kL-AtiuZ}b@-gm_s@aPn*KQ6%VN_feF41Qh@
zTIp4-AH<RJ|5<)BK|Np59T>3NH_Lkr7F)-X<0QJRag)+*alha9kDi+kT)**aM~OuX
z;pVthFZQg>MTz{ZIHyN2xS6f5X+Yll%Pc6pz_QxIryJ|1!(VnNx<VRg1V={xiOkj4
zQY-gN8UzT!r60-4R0Y8#f=!D^0VxT`I&YfO%bUn|P2fNb{<SI`3HKf<0~<92970?1
zixL{8FP|D@L9k5G|Lz}!(_-~<lVIkCyfcgcr8x{}OB>g|wCw3NJ51jbt(l9%W6)S@
zmC|YS1s(bU$_sHU{(l&L=Kq!9XJTRHgrb)*u{Cox|L^tx`y^mwW?*IhzcT!-PWGjG
zXtX$5xLteerem!(eGNAM@%!8t>#ZA-W35hp%uZ&uJ}P>=f7aDHW;ZUXD^BC#!`Ie1
zDm>Q#GR`v3(b@nejHDW5aHetq#Jtcrm>KoN^Ya^nv#P`MnB$=G;F$o)0j~gLrU1;u
z#K<_90_(H8Lkok`+n53tGB0;T0yCTQ!-J!Pi20^wHfNRw)^Pg{E-t9{E{+9m&Id(*
zW)heXARJl1fTd`w0{->gfu+US9R7;ha;OAw&7fRZn!qU9(bcrkfJ|tk0$St*sQt^c
zi26TyAo|v3R)+tnrR1Lf$n>wEoSlEe(kqH`U<Fi_Mdf95<v<Cjrt^tl7@AoD$0fd*
zZLh9}U*NIKug#z8Zp;B^|Dvj4|Ijgi=s)$=7Jm2<d2x`D`-cbDG4(-Mn(9N0-qUOo
zTT5L5UjLG1wXgrSKEvXd_`d;?|I8&o82~eXgIpV0Q<<9qfha<5tgm&l17ZO*xRLXV
zXay!_N9SMrD1@=0t>XUlw*N#WxUheVg7RxN_$L{E%Zv|5Ob_5#6zc7N#VwVDulz_p
z-E?r@^Sly3Gu7MIzs>O8|3mn<%Oi4XTE_l(AM;p$eF%bz>&v?Xk!Jp2?(bU5YZ`JQ
zLTWGtHa_raVR2=D)!yB-2Idd>Z$IcW{yHK1pZ^>+gw4dE-Tc{zfBLQm-s|rA5^n#5
zAt5(9ycQT8>VeMH(>VZTXsU7i^{V^4fBBtbbZ~G3T4(>zzVmc`v;MXu1oFA(F>HNC
zX4ARXnQB;@2i2O*YmpL7ZAS%COhX`@T{_P+zrsf1J0gA$NN;nXY<sbR2CV%N?{-IO
zk9D9M3~&$*%pIvRDYV@{-%5gu7pL1?zhoQsTL`u9poCz@M%CdCxeix3jf{BZ13V*H
zso$zqCkvwA!Z?(g1wG98NhamKHxGkqI8x|9ElXF?f$rgDVJEXaBo0y#|E7Mq)YX0G
z@djBjV|zUeME!pZ(p4W)Mfm`lLU%U>LEc92S|S}UP9Yofars1h&(@%Krv=f?%VKjg
zTU9T_H_ff8>=?4t8emYxOPKwNw9=e_bQbV=n<QBvB;Z4fF+CgbS2?CR$#c{YY$B54
zv3?PTnC0QSOJ;0Dl~LS?=^sjb-R@zpJLRV*-!c-uIOVEO%i<1$`73M6WarT>@{h-6
z(=HWd{*rZV9CS(kq0MH^Wuix>V?5(Dq1m7<onhHSqMnq)9s-h&Tw*<IHE@KJV?)h@
z@R`7le*O7|d_;||(aylV$@g2k?REDnW2IYVw}Ho00x<EJg-vJl1|kcP0fi6x7byKS
z6B-rOsZ{RzVG^VmT6a#LTeTdgS=L8YamPyQ<}L^eS2=JzW1gib<QOVmuso^WonD4W
z6^5X>9X~~@&MBBwUWpTI<B|&LUUojzz{qV&R5&1-4t&3WmJNSI<?LKt@m&}fi9yt1
z(x(fi24G?c+;9Vs{<hRb7nXpmK`2PrZ%e3FVIH4sOn1D-aYFfdZI=mlRFq&aM%KxV
zjFXc@G$#{n_K;buEejDT=)d3N)p@IUDNi=jHAlse;r~-?Ug0(yxhM2La%Uku`L=5D
zEHkvGW*t(+{gYioW9#5!;yS`wq*J-fuM%;5Hd_qcCW2T4Sy5c-k6Qq}o0@%qO95Hx
zxV~Y^4+fOs9iG_J0!_DtDA`XItQY0fc2m$xAD8T!%z~Er4k>_^oM*jHmz1i=L)DRE
zJ3Lg#HKj&e+Yd+v<H2+XWnpSG&+yv{<4xrK)a3Ekv)*(EeJINmES5fAyv_*+3|zL+
zHTp7OdQ&c*UyBx&oV`|bE6D1*r5)RRFauoMNlo>zJEIlWGKZaX6A;|9<yH63I8F+h
zfaR`SK6u_&D}GBz0L_kVlBd!=KLwITWc?l1PMxRDTQ=>H|4~#i{Ihipmm1ZA!~f%%
z43&(dY7)xHcXHM2293i`2y4^#)_+irYJ|*T#RY920(oPE1=)!j{yl?)bEL}IZUhYG
zzY#tA$)*(C9dtw2W%T2M*e9dVo9vDe&Ce$Y4W^1OayZV|{JbwWJfK_{zi_xpsDmze
zVv@W`zotD@V3!iza(3;C4?)oSW4##4^&7#W;eG@fWvq%yY^KkZQTD!-5#+xG%>}Bv
zj92!FbuXa9*&_&lpSxL)o!>Smo?%4t*0GNahYnvy<{;#f;1CA^w1S#FhYRtNu&?x7
zflNYq69|h2K|?Ye)NjE;!eJk+_kkVjbaFwtq?+vQk<*Wof^b&!4b0;0Rf+=}xs}ZZ
z`G(+Jf>;;cnu918EJ2^Ly6YYSei6Jkfz%GwTN*R8XO%-|JAbPUq=MenB)1@d_$6Wv
zA=I%jX{-F?LYW}z>O9-i{O1RMhAa)_t`=VvB#rh#!7&j-`BB^YsN)h>sA?{&ssa>e
zAbje`hhdG4emM0kh>99~a*IS8NR|mp-fVDFeP%4z_(P=t{J_q%Ua$|F3+r(8fi0r(
zbrEaMK~bXI1M9R?1>|kWjGwtsdeT~6CXL)oEg+MElFvbIsiT{#cno6kG58j{$D_*G
zKG)(Wi(AEh7V+Q*zBN~MPd^hu>~_+wb|&wpC<Y|{Un{u6YSNE##G`kGw6^7ih=S1D
zb3q9#xeg!RVY0Yn<^+EAOmA_SWw2@(JTv7C{Lf!9gLTD1p?&<KR){YM4HXiueT8kK
zr8!FR0WI*oLEere<~H=7U}%+g)FOq(x+GyE`4;UTZH9Pq+F3JhC!NSz^7w)Y_42R9
zq?-y(fpbu;KY$d1Y6U!!{;m=AnU(tk@INv-kj>6=<zcTCnips&5Fk7H#`o2a22?xz
zS_!sUZ|%SXbSdgO3}e0l2yqQzY~Cp0NSpRp#OF$$a_@J?kWDElRWL=;;Xyo*^B*y`
zU_{VH&9MrvGh6KOk>McqoRPHG>2b6HpW)A#kJ>x+QU^G4G1vN`WGe|b`+O3<o>n7y
zz3bmOT!D6iZn}l}>EK%N0%OuOV4XoCprAg4(0Ue~FL$zioh(2D08|!?rnSsUtp*Qo
zonsD1Pven%8Q{_7PjSCAn6kSkwJMqUi?nVESj6n<3FV)MPo|`8H9|dP<kljDksRTP
zXzKo#*KI^+i_Dl(zH30Y?~nmq+uJbx`B3wv3D840t-pT9t*4Izcg|ahj5xCfUrX8Q
zObA!7!U&gRhR3_u@HTcTFO$t~$Tj91(iHUPb05+x__24~fjT+H4)?2k3K>yW&=+JI
z7voy61^&R6lraP8+q#wIVooBz=~m*<Ziv>MsQW%Ec1dg&_6YXS;)i@R_Mf}_T3T3H
zTPhJqXbAJyV~r!tHqj<fW3-oH-$5hsv4$j=SL~g_iM!^;TmSLhadH;pOuLO3C>{l9
z6a4s%<+%9NMz<BYGu`nn51nvO`%Y!IB*qVgo>who%m`R;+8y%ZvK5?X%(W(IpU~o-
z>Yw43wGF4fKI&!VwgnPF+tO@KvQXk2(>xn7qf4kFRapHcjV3A74J9_A+?H)@2>bp4
zdSaL0>P;Ahj=!zy3T(SR$#EzsSFhvB_09X!8W-58SnnR(I2JW7s2oG?fsM~m#}x$K
z;1FQ`*x$4<1RP3MpF*)?F>`?oXOwDhFH!pVM+8Q5ihi$%EPG_?5q)p9Kc39$fa2gl
z3X%zNpkG*eavyq(gS9|zINV8Vwo;u+kI{vCj#04WGr?Dh@NE2?6e#zjf}tZd0oUa3
z3Rsb=3sWHPRPqWV87z1_ly6{=<mHh@Dm12TUv4mqz|#Aa>XhaHO3Ub3Kq6zX*2Tq8
z{t6()yBa|l2MVzLNh;(-U(F8bMT_H#ONC<Xz$Yr?S%^LT4!c}>+imVredgwaXyyUW
zDuLFk`sEt%+t!Kue>R6E66MCko-SihNCQhy7ZS#NgT57~Tdl6WeChv^!7y;l@p%bD
zDm>YtBGwk4kK2RWfv*PQwnJmp4E_=#!~G+*6$cz9jhgYx$Tc;+?XaNmJuNMZN3*fr
zkuohl%|<@<g4ARka9RY58Py@SNPn1HAw%)T<y#F35scEE@aAfBy%V5)0nvC$vQ*?5
zoQ5@Mg@m%&=$;tinjwowF#D9lmDLc!KjsLDdBxYo2iT+e9qiD^zSd(-U}3y_r<54p
za)g>&U67>r^1S5S5uvIow7Kr2<i%V>P?O|DUW-=nBD<9L@tiybD8+rF{I-}3x5u}3
z*gI|;PHV5S*kD00R*??@!4kS7tW|JRYyH&!Yu6q}wj9H2afEkx0d0kjP;>&>STTDk
zoHpp9zYQ8*m8a5cki~8Kr6Q&UEL|nhTaSE-@y;Q(miKr4u2ZS2t#^Huw1WNVHq=Xb
zTe$W_Sjcz!fWO?3ye>pfTxFe`GBM(CD5cHela&Kf?yh1SBShUR+w7tMw2fkr*7!V0
z7b~J#jwNcJtE)2NY@?`U<ER#aa~MEyT?8P-i1fAS4WIq+-Zw0S`nuyp+asp)Ul6}S
zNpC(-#skj6p^Ic9p8O$x{Ptd`ct#R<<sOPz{iE&^Y^<=w=`L0^jCnrFvNw2q)Legl
zy<=?21;$tJuNm4kUAxLnvk0B0f;zY5WFEPH&)*sq6C=MPzyCK=TbbCZq|kh%d(!)m
z;SJP$;#lN<d6kQ_dB>Ttc>0g*!g;qX(0QTKpn>?zd#tj1|Eu?axEwTI6w<5JzFHJx
z%J*`a;cb)|874@A^knqerJ2e#IMLI$_I}!}3q<>7p1jIP3U*_sR4M3ZuI)KJ0eV|y
z`1*=dw7g8A4=tu+@+{oZ?+1gNI4kE9T&wia4R3^*sX^%EJ+Zb>TuQX6EQrub{`9(3
zAS(`C$)qN41$n^oV#bHYt9K#i*6Ia;cK7?@TE{KgmoHess~z+3Eo>2(>p&C7uFt_=
zTmd#Q4L49aA9gYRW#6yna7(MCvx<3jlj>;2)>S!AQ2Il2>L;@m2w9fsXT~FKgf0h0
z&oCFfXr!o2u_6*O)!<E9^3aCWdxi%ncQ(i5qiv1wjC)?^DL5|cW~~*|NZJ?9G<)AH
zy`U&5yN4=?Tt{&9*87b-IU*>8k$EFOt>eWI&n?)*rXj;yJ^KJXg5ehyuq1hHJhKNN
z`aR=H4E4fHR&(C|!$<Fn62tPw4g`X@tORZ2E5UgvxIEmlqV!?V`&`cvh}Y-dJe{MO
z>}SlaCk!M?*lKffj3ToofP;-0XSPrHW;cn8mD0-Hb7Y1i>v?+QNb1bH3nQJjwv)_i
z!YgUeP@&TGoTI86t3-I>7|eLy?oOk<LEc1?^VPYDbCfqjuy!hP^3KO@Fk2$BRf%;G
zNbh>FFq!`5Dg<(U^|<^P>nrd63R1uot4#7g56q4;5N)blP*cC)pZAO=aRic&I>?#L
z*RYv(^Nfx0SeAf5u`MXo^h4gms)|pd<okLjh=2<+n*Flde^4*=GbM>(l|JlNE$)8`
z@-e>af&&*w9T#)n{OTK%je1)r_o>}wCCUewEJxxyd%^D0Io{_?>HGUcCps7wuE1{t
zk0k0-QfL|Pg|QDfN&Ym#RZ5zCHFc4lZCFhs=;BY_Qon_tt!73u)6!rsD06MWopFn=
zCMfiE0N@r#9~!N5T0b9V%98lG8Y|8?N>4Y&ux;4X@wNbQtv>Z|3#ByflGNM<h(>Rz
z47-oPat)Abz!_rHC7o$!|36{(N%_{DB^pwB8_c*+d19V5)72-yYi)0c31)-=k^($0
zI-YiOD@i5PZBCPPN`1fKO_psvkhzs`W%)Vq0fD*3xNvP5168ZpAnd|+_%}+ionEZi
zVe%6*c(qqV71<jh?%uP)U28%{r1++ZTAtTf?17}eBcaWjdI}8=_2t(Yc}Dx&R;DFS
za8u7Hq55_R;1sxoSK%)YJXJARYGd%`L~^$DWa6VY4D<DjtXYI(xc|a$0l)I1B_C*;
z7yMPtlK^wB$7Vcv{`OAPj>V^PxzRQ1m>%pXpGhMZ(jx#No%!~T=K}NBdC79REA$M}
z-^}`!#1+r+Xpg@WdP|`JU?1;w^noDh&i&Mk)mMxxtbe_lp1M`D-M(|#ZF`&PmwB$*
z{Pv|S8n4G$bIfqPQ?kShjXBF}xLqJTSs@_^P*LGnEvMM|c)ElyYoU<x2*RFqPpObg
zqztWImySUE<NxV5Hr^|p$QDxtva#$c@x;)~Np1C^+Wqv;AE9mAS#(@svzJR*x+fSg
zypZ#PlFr?tk8>+)dyIV85N<)dtt24ufM%D9ry*qKL?8CerKTYCN86;pupXF}dT-&K
z@7+<q^y#$I$)*j!CXl$bjOJj9!QLo~t2heD*b^PBy(v`5Wd2^}dHThcARP@3z$Q5J
z5hF<ojiBm^8X_9r+VRH_$mhwSsx~<p;5khs2$c(4MI@*!EdWt{L82453jvSs97>)p
zN9ND*%CncKs9@4$o%oVRz`y9ttq6zMJaAg?5f_A{-8><@e)iBo_zR4^1_<MEZ18qi
z>5tzrdLFV@EkMLw4n7jw?%4Bizgh_W6bVE~9beg7`_15ny<L$;rTy_Ji@}-vZ-nYy
zbf{-m8u9of70|}tTqKm-`KavZxE%qtB(q+ln^%47nbhYasXHp?%w*zN>2H&0m#ZGl
z>55VGIIs}rdKLrKo05F>eQ@J3NOIuYu3m!nNXu!XE!ak6ysuuGv&qwOT>{JI2`q{j
zK|9Q;Y@08xA$n>9_@t@dTB-sTp6Yel-HKSF`R?^tmXW%5L{>(c-Ba_`=lM$fAYApM
z?s?t%8h#jN=7eIr-dg)zG>VI`)gI~%xcX^Zs9BU-^~N<oG$H>sY;EU_EXNtVH56Km
zlm#H8W*)+Ph-)2uxXUt~k^`WHU2e|d;&V=Yb!gHelj#t-*xEC)JWm;$WqxOV!+3;(
z^GyM%F>?zu<v2sFdhj3c(t-pWt5uus_W!je5G!q6uL3j|2uRTKmGUo2MRF@LSWS=s
zVk&M_C1o+DA?pP3V46`B2m~fBW=!o_(7UH6z!*Sg?27ATMU#!0?2xWdH=#F{#_k6a
z4Ii7x1o$A>sZhbb9fK=Hm6)Ly+fa=WkNi8a2UBN!*<Q2(t)Q7g4Rz-T;^QMPl{DeK
zDDY00bJ!&VA!T}%oQzvf-ZMg5Ax{k$qj|S;WH3IIWTG`u?l!bvNQQYx*ZKf+Hhk6G
zu%#6l;BcF5mYa<-qpwM~T(ViM=xPLc29dU)`28m<jP6q8gDr=x^~~PEO_3oV8cIeH
zy#pPIhjqj?(=EmCR~c#L<4yNPl!pqyZJ{&>G|`zz>+l=<t6Caf{rFW#9D@M5GSzX(
zL7r4$cbDMdI=@yHLBRhg(by{dSa;)PR{h%wv^G=^V8_;N*l4$7y@IxD9J`Hik+7G|
z!fdQ$6gNh{s;k^tP<J*E9&`YCs)OwG-tsldRm=GyXR)wdW@!hn$HYNmfgqY^IBKDk
zf+Ps2?0%m<qe6`!R(Z-&-Yw^)wOeG%LvskYv0XGNO3@rsjIAs0?kkFCnFDqpBZW^Y
z<y-@5BrM}^T4{szBt#NCx0?my=8;Hp_1a6IO!?E|v5l!AALTIu5uAU8lzox})N@aG
ztN6RInWyjvy?w;MJkMoGC<<OQcF_PXQ9abL*1)HU2Zge3FI%fK0ZsZ}5m){0pM`F;
z%PWU{wvq|{#q6*vDe3YEDkXIr6q1K58n3=PHm|`scO;>|V+X<mz@5O@6)L>n%w|(u
z3QKXc2sP?zN5AJfV1ye~)3g#ydg}N%`uk#=!5oyc(yrAxO4DiVizGciI!yZzmtwKO
zr|0@E5f8*8!nM{KEz#V(Rsu9#e_$BOnLH6)3r~Ct(jf(=w9v-qYc(b?>#OUTGmO*+
zu1(+S=4{O)>-KZ&K*^mW(oVCR^#Y;;D(<l7T=rLrw<xHRqZlv~^78D+OEsb9Zdklt
z8!roaZ(>`zm7?<%S!JL?otc65w-`N!4D`;mEw)0Xe!%kGKXP^4=P$mDWJZ!Lc0%D$
ze*g`Xv8^nx1vv(+jDQ)_K=?EQZ{D`<4S23997LSPKK{H#>U2#&gINT%409g7H+h4y
z<dw17F`!wR^>NP{fhlkNdC;6@3C37y8NdeiA<BE5E~o6&r)pcmCFLxFcbEhO{qMCr
zqWlp+vWTD+Fkk3EBYIYg*>8qi=b$)v?dK$X>HxgLeI-_Sep%o85%}2825yQbs%~s#
z!R)o#)-giJh$iRTl}cRTmmNsx-M0zgE1z%mKJR97Qgj2I?a;2Sqw{HVum!izbJght
zP3Ymw5Nt7D4*<~d>tq(4Xm)18v)SL)Evi-Et_H&p^&%skZsNDxiHFn)ksPxJVpq_+
zTsk?d53i{w%I$f`9u%Hj+?ZRn28{Wrbw3%VEHm>B*SgNKw^CCDUbs|$G(4`E>CEp$
z04j>Iz19f=S}3%Pc}SH^egc>J?1rYNS)-(Vn-KPh?ilQ<6??yOYUso&fW$PGwEu#S
zj12)Q6%Jn<R|1}Nu%WI{dE@h_s{WM_VeB-gJ+_-Hte2+NwZWqA&xa);jLY#RXnfq@
z_YYw;l6<Po+4?i44Z`*`C>n}WM-83db^93UKt595kVUu5MO;R>Aq$R-rEH=)L(de?
z=tJ-9eHRe!EG$of&DCIE=*;Wvi2$JZ09-D(X-l-Lc#ak@b>lH_d(VH=<paX6%~MSf
zcUpH>GA~*$HahNEBw-R-N6EniBV<U`0`i@;$Ml7=zPC@r5vY)YGTkcSM0ei2#TYy_
zt_mGhI8ChOZKoDzbI&@IU^@)bx~ht(VD}+}Q}lboTRZR2XS`%W5J>@BeEfqqMyxW~
zPU&bA9P_9s^9`YR4nlC~O$}pDPe+jFNkz(M*2V8^>quTA?D;c1dVe2YszA|AWw=*_
z?sr=7m%9i^{u}$RS4{=5!A|7MhD{?aypc4xW<AM?H=;8YWL$<$Q05SWg5&~h0W+li
ze$tv`9S?pW{I-Z~<_uOYAL+#q3Z(xC(0y;8-?iHLUEUiz4%foOr&X9GmFY;UTT-8I
zGsc-pDI*}}-}V5J&p;^oKjqPU=1ywjMP`Ic+1>YSqN&J3fp7W+1jXtfW3BrnF>{>|
z^gIF*vq}~{N!x}^*{*Wo0FNjgkR>q(0wR>!?W6neU2-iX3G;o)`v!_D#-FX928GWL
z%ynmJ8$?`ZVfi&-`E1Pn1n`bUM?-KBT=hLvcRM-;cZ>Uem@{fFncAl;u9gVNQ5C6F
z^Su8~<^aChMb1XSAc;+OzCfaaG{!IN-S@>1C%}&95*Rxr`=7oFt}W`fEWE$`|0?SB
zJ*jRsvwGys3RTv!b9Y@KASC!Y1r98QSfyHq2NUF3SX;TOK*Z{48Y&c85a%E>Lo&VF
zczcS7{#rVzA}#u-;OJGHVd&fLBo^<Ue+d+ODbZ1mvz}90)Yb2Vu}t7OQ_EB?7>?|i
zche!clFuquoxd|U??PhO|Iih+ooUtPzE0x6TM}~N@Gm45Zj`NuauNb--***o&vBBe
zLtGu}oGH<{)2P_POgTh;909jK8hHK)$MTzRDfqVL39A!kq&!~nG8NVu<R$&Ik=({H
z8ZV<FQH!%xXyG+5!6(=Fa#7hQSY8qnib9(!Ls<*@v(Li5G~1zzTvg1$75q^Yz+7al
z#fb3Tc8m4@F;7j5+vR1NEttC|1FB8bEAmGtg0jQ6^4O^u-`w$Y?@ZXZlL9J{s?}HL
zWYjJO36Oj#r=@~lnI7V3H8%pf5N*&QxXj@a8nKeGeY<yv5!#l)4I11j+aTUZ5|B&x
z0ymppWv87WoCg)6Kd=UK4)-_A>5U;BE<TF(7)xd9DevTjH*4X^bw95+)5SF%F5T#*
z1Yj7U%D)jVshZB=D13=DlqlJ#>eNJ8>yhx`<da0iC-anzlL2)4{VuNjS{SAeJUx1`
zzQKG(N0DSy+7pKNjEY#=U|?J*<1)0nN-U+JC|ZJ*XD*v8Tw=jv?+aN>>{rUeakBJ(
zQQK7)KeGi77gsO2I$A7ot1K4=k&UX(OJq57;z9MUY%uw<*iJ6kyV=K~EZ+pbiND<6
zl7IAwA$gTEWh@DeKOo4_OQ<79N-8J3nq$jx@El2*>&47$@H+GWV!6Qxv)*6Tj<7^)
zSq`UL3Yxx5B?`&RYm(1lUKTBM!CfeMo~AlsN3X=pg5Dr;#gYr%K`D|e<<j5<4%{om
z%vvq_3{G&NIDr-ad;_r{sG>0a1N+-0`$byWwHxX~0A3Yx&Vd+wN7iO>6P)dhWSd7m
zkg;ZcxzK;s9r$lrxCbBn&S#YnWFC2pLtt)%{<6G0yAkj=DRc&3@}@4-GiI#S6*1*X
z=;7$|r*>z*BMmcM!uxgmoWHJVbW-B=*Zzt^wlSAm$j!eQnCpCV8+od?v*xH3w}ay+
zAzQ%Qr^G(}w2368xp?5efl@{C(%rvJ@8ji}ieqfwp%$X?GAB<O%o{jJqf5r0>XmG$
zRyLQOmfaYGT~D`<9qgj*-BMTR)s&J_^<8bR7P;R|m;>{$`IA#?rsS1G+7@WkkkUoP
zkk6E{UZfI{X8B{W9xT&2Pn2?s_#&F~@|)-9FfI@fKiSAAD|Q^-uz_a<UMIRtiErzD
z=vKOPb^cTvqZ^2bLHJ|oaua9PQ+wNvCN@PWhD#apytMlLszN;78nf)N;Dxs-@q=w9
zPUl;<p+k$^C_r#^NKTUYEuy{!vlGlH2saL7JB_BtM{&zuISYwKa$pk=um0h6HG&qC
znGi)y14VpP@u|dXL2%$M!Fp>QCXj_JZb3`WA#|jlV^evDIoLX!rg_f*uj%EeJ&aFB
zg15#tB3fE`IXtZf2wNn*^jwt(^r2U{7k_~$@v}$>T(RIyVE}f0`XUKHHpkJ)#ftXR
z5yHqg2Q)4R(K!WWyOi<rKi|1g0vzUwK7WX#I)HycQ{>RpZ17gnZUn^?;KuZil6AXt
z+eT19SKNy_o*rm5SITMeMM91dEg5cG<$H^5ut9P#XEa&k`qokVPW*vFhrm*xPoQmM
z5;sb&VH4Pol^x35ee+J<5nl=1kLg?gBy8I?PAd&0ze{;dp6dJh9ln5IKz8&jUDZDP
z7*;BY?IeIpG@fFdX7GRcS}Wn7j@+?uM;)eB-A4#FltT``k4#5wm)VJyA2B{9r`_`I
zMrOsttVVLz?j#F;4bSD^;_|kco4$bGR!=%?QoL`_7^2CWU1mU7><{~~fD%28dZd==
zA%GZEIl9U+T2V(nyB|q$t9_2vuMj%NSI^`HV4)KMQ+Dq0-)<l^NXg~7_}OkIpwrWE
zn*`t_t&gO2#9z8`OO0`8=nt_PIL+B}v2mK$%x300d38*Tnc6ec`;4rcSR3KlchWgh
zrkLhT`rgIo{0cwu;HS;54$G$|Ti6fgmZm+eP_<^6vGmJs|A(=22oePdvTfV$*S2lj
zwr$(CZQHhO+qPfZ*1SIx6Y(cv7PG6`*CI13&pk<v<yIR?q0%boRkch}1vVzKd8Ji^
z%PsO*))Ng1?t6pD1a<_)d=D01E;NxMhJUZ&3!pih$EkZT5^Y?+O5K%Ddle**Tu!k8
zm=L7=AxL*8N*3{8TNcU`sBAtWLtrp=?LDk8nIE1WNp~n@?&HK8TgXl~ExrUVqPVf{
zxnT<Xk#<Bqot|*<!ef((ZQi`!TIi<lDmb3r=Fy-h<CA0uSh6f&;<8|bXh3xP*oUT`
z9i06FqV}iM>Y_9mXpD1qW24A00yxR*vx^0s<88ZOk;9?oP9zPqSADbdyw0?_eMxrx
zQ*4`RK~Smn!|%4Qh|;{hF8k9RbCQ=XsZZ3Edb5qg_fRFfB(U199mc1n<5}3t`w#<K
zrN$y%3G%KCmDQHI&RXC|Q=G$d%T-^;v8}4of&TVsJmLSSt_emxq6vQ+96omeVewob
zFv#Hr9=&J<p)0BPFu_%>_@25s1aot0O6f@>FG1XE6S#AGsQr>D=Q~00?A!MbOIw47
z>f?ksevWr*K`<3G=?VKIk^tx{MR}V_=xPc%3_F<Lv2YK$jpSj}r}-$Y+z%^oXKPB{
zMz@}vS3ffGitgTtl}l?C3xew5OgWY7H&~u}!HA#8u$!_xiDk5+Np<m+;d@a}L~n6n
z8A=$fFSBIFk|@9w{G*GQuj8{ss0vF447L~A$4yNRbI<4t54xL>trvdwfXK|@U_3(Y
zdH;Cz-fPJNhnP^hBXyLYwSeQYUK3^Zuw2hlTnJok0#piXqmRvk73t#9$@-S%IxZEb
zv1u{@*QX&%@(I8(qT(3=>h)(7aHz&sA{Fh8&B08C3;UDrgb9-dXa#9xg9vK;cvCV%
zuXo-RH+1wU!{YqZLABT;!k+ktJqcA=QKDPF-%5Q#Dzi+k=tG(}xbRl%N5TSQce-b^
z2EZ}z@+-loF~o{FJrpQ)+`aQjfs<870Agb+dg4N4iRL`U$jF@u9M7QDg$UcK`H+gh
z=lPK#d45#}R%$DT-f*9m#5R!y=ITw+|D$y@o_u`6K&a(KuzORrHo|a9%L(jdIi2mU
zTZ8oqH&c8&de`s2jX7mKk14A0d0o9%FGQ2ZOPvzrZMb{^ba6}=U}_e8fJT-lU~3^1
zICcgSF~{PP$IQ=(>U2K628z(=T%SlStI7COyB`#B#E!tUY%dE{*S+?DaH%ve<%I;V
zm;sV_QeD>(U$uvD_fR3%Uge^rqz0BfjEx1zMwOxCXIe!@0@e*MEU94FVI^uc^ls2+
z#&(2F=E?&#yi;xMA<q8+LOcBhJHRNjTyr29YdSG?I#PN?S2o8KowTSAg9#zs>PpAV
zSt_nV&GJ3=o&Q>?2Te)%1HH8wZ34a`Gip`3>A7O77v4&ydT_}l0v9)Og*Z1JAoWH^
z3@<UYsJ>n3k)3cQs^_}lZ4BRVg0Ggw0)FDTkax(mkas28U7RI6VtD7!=IEl<rBE=0
zb+TnhO5{@u_uXIg1KISSr6zqnHLWJd9h~v^T%zGZ_7l80r|B!zh-cZ;#wp4gj3&nl
zgF9uq67(<sUSGA7^L@ALgE^|kz+zU6z=~!g`c}p6Zd{(ZQdWSDX2PhX`pA7Vs)clq
zc7xPZ3S|_Xj8L1z-9oiO-XM7==-qSV&8!hnP&R>qeelGs{BC5@=yg~9L<lfe+g<qY
zWG87@`rVKlO33xc<7=r435~Mrt<sQ{?4FIU#zYw6CB#tpBhw7HDsUr<-@`i&dE4o%
zz!g(xP%)OSCL5I^S^EM2e4esJM;c{<?pm~f4jKgAvW^$v4pk)+H)3X`si$)kN_LFy
zT0nwU0i?H)bZiM24$n41kxZw)<M73dwBTfxY;aVvw#LBrx~|Hzw4DvMwiYw$OXLFH
zEo+R?@9>TgT@!u^1l@HL?GA2q98%xgI*7t|IO!KmG41GMOz9S+$A&>s%ecj%I9{XI
zyw&9_sOgq!5wl_%WLe}h(4iH-XeSl#I}W&*VBL0h_s|it_%?p_>PpAUWU{R6>k$YZ
z?U-Ki-_zoDA(}UXRDMIKKG&5JQj1)zp&zSs|50Ie<qnic1O9-GJMZWZTU2Sapb$$o
zJgi{xE_cw1YyYMT_vNtFfGJ3D<rN0&B;b-w=}hvegXEt61(^iQu;O0V=G9mV3esWM
zSXcRDY#vy{)m$CkOM-<x_dq$ssrLM-Q5(J^wr>t+CyaTVHQ@}T>1@p<lP<Me;giJd
z)qg!&4n+bNZ@CiJ;W>WA3}B8>#c>D+(Y>1ZT`{R)4aX5I8|9Rl+kp(^>G4oREM)Uv
zFC5ViS0u57QSoQ}JO^gP9N`>ttQgQ@cFZ?V?2cZ{@m&U=aHo|FB1>{h1mc!j>V4T=
z3-|rSu-qM0_%vp<<@`@0L)4um3y_DfI^rJxhyFutY9_odQMPzA!~vZJx_$87z2l13
zINsu7ewAT@4&PAy%Ta=n&bAqKqz1yEnI`v)dWs48`EHL99BUTFZpaIe+@;zL)gR!P
zPYg4<?&U7(-;uOUzn-354SOP_`K^5r4e;LejtoOKJB-KLnC1BgYo|9g1H}4^9TysQ
zRb6A(#t=Vkd<<Q9npvc}r0$T~s&JI}EJ7*x6ST#in)bvRk^bI5r^l<VuE>KZ6j|ny
zN_E3ch&qeQXt-#14T!piZ<Fzo5|ma-Mk18<XP8c|J4{EM2RKZ~!ieUZ^)?&OM9tu+
zW?DdLR?1#}!^N<=n*9>04h%75JEgwVc@?L03t3DYE<`UST-LvVMK>b!6(HM27<x@V
z2=h;CXowCa@zI{H-oztx+cdtOcU-N9+a9Ib2iKO*3@171*qCRkc3|{SmE2am()0bR
z&+J+cHII7AsS{rz_Rl2CHgHkZM#<Z-*C)AbB~V%(Mql6O7E%@;#M~-H1r0fh{SIV7
zL%EmESZs_8$%aPG0~`y5Bez@iOSuBGmoouKO*W%lOjtFcC`&sW`WYqB-HHW@DC3m%
zM$J-=>;Bg>Q#H?&l#1Z;Z6S2po?!N<^={dl8|^uhD#tnlSRAi3a14M#yfy`asRp<=
zpz-!xH2H?a)$#?-9JEI%B0RI?GK7}$l|!P^#y#Jd_BHUbI*H~}3W^zL?U((h@@2&l
z3QhwEvC6kxFA6?rN`+j0rh91PGfF#`{vh}=*9;*~i>vvn8ckE@5=9DheDS-okA6Oh
zXt_``Xn7XnBzG!q2SkoYC@=eToJAy);uj*yUiU3yUh_>L?>rgvX0n{zMb79SMG7e{
z7eWOUydDyzfM4*m05hL;>*zy$ezc{G5&jDWl{a6emFZ<qvm0p!De7HBbgSpMPEX~y
z=Lq9$k|?d*8Y&zI$>21aPH+7Y+WLH^k&DP+9$j+cyX>|Xt9S_Cl_ybt)W&d!H5jbH
zc?y=W*&e?6pFnBhSc%;PxxY&KnP7A69aE1!fR=?FUVeKq26Bd?!2KvX<y%%BY)x+>
zjK>J+Mw8$+**c*(y!^d3u#&zE4TiJDwmCO`I~L6y?j*x8Qh+N>pbnq$!`o7HK*vj6
z99NyN$f;x0`b*NlKue-9pyrMROv-aF;q@={SK<EKdG`gxBrKm{`^I$ygG4JmW05Tx
z+Y9g>8QVv%6FER>EHb_`T#=1%gR^lDC5|02v{oDo$y<5WTiuQ(JE>%WP2I>%$GUyn
zE+Sq03G}K309PHT*fCHP9HIl22G$(RL#QT(G`F781wXXU$L*8X3p%@4%mqFLi6eC3
z)<a8k&uf*f3?o*;x@wCjGC?xH9$`S7K03<`NTdU6s)dUM@Gh`VZSsB!{U!hA=!+o+
znnour7LT_2ZEU-j^{#v17Jj}VsF8ak(e*<pDT$1sa8Dy|#B_m*I+ETHbr2-XN_Ai7
zduR(`6nA%xtbN5#RSuGul%Q#e4};|AQ|loD9kXIQ!E2Wf4^Ee7C-T^7G;t|nPyw4f
zg)wVoIMA|lfxb*0>J{W*?Zv;RXX+K95%}|)?Sq6e00d*w{FFUQiXH?mBcB>o*15`f
zuMf~s5rWif#YOMnGWR}QBDqUNl5%|L(ZWe=E9flYJE>>Xb;5y*m-Y~RJG&#1gi==y
zN^;zud&Upmf6XrKCHwE$pzv=xCT$)%13aYf-Je=w7VobNQB_750bS2x9n3d=a}A!_
z;yd*@LShBZ9XXC%PRf~Ni`t5j&SHrbs6c$^3a##(8}C4!0g}A@yi}G^Ke@<qD2$cv
zDVFT-RoN!|6#`H7faA^ac-Ja8SN9+hmhYgw=4b4GEMUeQay4QsesY??<MJIygBqVq
z8n}_6Czxem6Oa%yERXb0|3W#yd_~4(qI_I}CduU2VH)3`Q{oukN%v)FgX-#IDV6*Z
zW`R=5IZQIsQd#V&`aD_QjFroAc<{#$3`U#ST`;qLaa&l4wGJGGIy*m*@CM~8IBhP!
zs?D2dClTzI!CFsu<gUorQ5qDmjWz#;Z1(>n5U&k>P*WhJ7EIuN@OhHA8-t-$!dO{-
za}vYDdk6<*+|Tt4QgQun<v~kANr=gB&jJa8M@d~2o@&QAIc?;2-Fp{1n2EZg#Adut
zgI3o@#wxnQZ?b;@9WU;KC)*k~SV%`_=%hbuM7h=~@k2GL_4%dBk>7DosPRvOf<jCJ
ze_n(vndyYVY=4>7R$4-_zg*179}}`qSe?<Paq8plqB-H(*IIlsN8WhjYI`rF!qMhD
zQCxF@mi>(86vUH>4i3adl3WIu5OYh&@1ozI&~=@1u@rXRpR=jRuEE(7M5z^!iRqHs
zlp{KcwIkTQJen!+$q1!ZWeIbaJ`g{{6v<w0R6<dT78v$flkz_Jl-Ap;T&^qqq{EzA
z*meuz?OEn5AwB(0rrE~Ytuzk7MISH2KMsgucf`UHJ8t|mCZnyjnc{x%>#pCO4Cuz+
zDA8x}jL68U<+PmTir%5I-v9hZEKGf8`a4hQs3~E7iOKTXr`{_BP<a}?*xNS68wm!{
z8<Suj7Pbta!B0&h_oL`BcMsjLe6d84I)2<`ZC^Auq{DhXzF3G5T!u%X*R$%xqw?UZ
zE#~E<qBO^s8?28vDjT8*o7Vy2XS;3+-2WbpL-a=74AJjCcRV^(*^Ov)H^=M25lf;D
zwZ`^*r)sw{2o^=BV{38YUoQireEul{)s7PXHlMsCgBP_zU-MbYzsG1LArx9{lq=v&
zfIHS&LMu8`6P3XIsSk|D;yn$!>#I0oZ#bzK`_^SX1$OU;PpsZ#41crDj@u@tm>jzl
z?)n+n(F~r+Zsr<tl92`IV5<7ts}6SLgGnk~jf}-u7yc0c&=@@I1A$&rJmt-^>bW|_
zE;uUDK(KdVf98T5r&^bo9rDO`Si&W%5`)aH2u9q6lc7f(vPlfEF@fc+5@dd~dwXqY
z)5I+Yu|sJ>LG{d_!B^!<yp`)!sxX7`2GOzR`8iF2gFzHb6Vw@9=-A-pO}^|#Yg#8V
zET9xVK&Uu5jK{7WsUAhBfoEM^D0tN`9kZ8H=|-$o2(8%0JF{u90!s*drVJ(Xb7q|o
z5TdY27=Yh)QA`$G0ou&L<bf=>;~F?M`KTC3Ip7d?r3hA>3p+{Eoi!<kfT5zv$ud%!
zA93{>opKq{3~RFUb3dxeig8thZIN>;8fO?H%yIby(^+TeMa9p&RnR_wPd^O(ooL_l
zg`q6##^C$R)pCK@wgVD*XFG#QxW++TuG6a2V@PFf==eag$gvjmI<>%*3L{FM`)iMP
z?|tCG%WMWdd*O-Dq?l++Uaq<o1@v%c=7lQ%!(3M(&EHjIrH?3R^(csA;5CnsX@j+f
zqdh2LGpf5N?VUPY83&#G{w_C6(-{O)Vi5tExIk2UaP_Dbi6-=mXE4hN_P;PUEdMKW
z!}{N34I@4yGb0D{e_#K9=7y1(o$3GT&9v?`iZ`8UrO|GUsjV!t{!epe4eizy?SG!0
zd%SMFUSgb%qMpqRjZf2w%&Ji^I>aGIuCBF}xpAa1EHKW|+W{sGDHEe%U|;~iI8D(N
z6%ELggwUkCxDbWVlxY1sD<JtdmI3?6LG%sv^@9mOA-1^OIW;o1+5yO;DSmN4$~Vw6
z)i*Uc0hh0Au6C$nXaKlvZEb;UXKS5hYrUO+VH8<eK-<;Rfu^Z#qyYyXpfVF34*{SY
zI8^-GD{?B53SjvMR>nsc(eaHetS!&1B_ZjX+<?}<#sJZ`wldUzqLP?)`XK*R%$yxx
zLetX|yQuh7m4xJ^bY(yYD5YwMVCb6}0LH{W7`A7J)6ejm{}R0)^i~!Cw12EN(0|q#
zKkA<Ms*}e2s5#L^Mg1d68ld~fm$w(;LVxr+B{nlQfY*PJSsh&7@|O96vwsi3^WVgT
zB>f{QHu%idu&_9Uw3Gls>U(N}s-h~8`B!&`mnY!x&kf(?On!lxnp>LhzkWe~@P_^V
zSzJ$l#l@z^1;_dy^4d$9(o4cBA_7aM_yV&qHou6L`;IP3e>ZS6l#r#=wtl4zKjJ8V
zud4n_o~iMgXZX`Y*^}K|9+?RC?6<YQ_dn#Bk^d#ny#0UjOg~SE{yV?2&$hWYKOYSt
zG%_i8e^%g~zRchQy}4iY&R;YVV&mHbnW+h=d=pbc5c)>?Mj-Z#4DVkwEh9BDG9|7w
zzp5WOir=H(t}JO89_bi^)zs^l4x#GBe_o_7;V9FofeN!SSmlZ|i~zOLPP^s=8gLrv
zC-^%18x{M9or615pcr>IIyd6mtZ<kk>4R7@Qa@zTzqCi)KLoUFuT{IAYSivkQS)73
zN<*zIslZ(k-A)tgdeDmbnZ^^7Jaws#)CVG{W?we)o4eq#%{YCXcjjcW6;iQ@70<9{
zopL%M%vOk+?l!}}oo!%9xp>{sZbFeYb_(jKf9=t0J<2)&-3UOgCbt&=9IR#Wnp|*3
zY;s11_tMW>e2@rR&D$wk$fP_k7}>jj9i78n`(x_0A?xm1ZGb-9FIgl$2K%^?u__}{
z(WS>*QeLLW4Rqf%zqgj$%2f?V_$EicJnEQ4g>&Omcs2K*n^Y|?N99wh%)Eo?m5KHG
zQ{dtltILw-q<t)55pTZn^J#el;NES&;{+@{&@4z;p7-N_^q8I50rMSADo7E>F$XiG
zVR?Gu^w&5UhsVALOy~QzhUe=q-ET`pClUP`f5etr3>(1VlbfyEiK?iq2jA2hbWCM)
z06_>^bO<Rm=Mb2bT^D_mA2)Q^ik%`Mi33kZsFPA|tbCN)F*2zfg`~NE>#03K6efA2
z21^8rsv#U5xEpdenP8|KLrOgWNUCVG#RxbBSiu3G8TzXJwbI}Q7jONV^|mk&l2q{#
zwJM=T@DTqN-|Y<Ib|SI8a*&}eIEBfDBs;*$2S=h$?G8K|;1kY?6~=1$PL|V0PbKa2
z*1K)yHp=!iZg3-B8-)^#pa)%Q?2;G%_}Df_d94J3y?d%w+r{>_)7MF|fFJV+0sDaU
zz3}hR6obsyL9BF<t4r#In-C&N;K2sUKzMkE;|{&~+*D>I@;ba#2qV1%D)UR&Z1KYt
zv-hWq5wuo)O=Y0+z0%_p-nv}Jfq4Msd>mKNs<1+LAoc%<==!`gRjpZ7R4S#NgbRN5
zH{7K4k4^uL)e1es(lU;Sa1GND25*(EZ)>7#&NQ}p@b1d--xZ|SrVw}^>DHs_C>Toq
zOpwH%&t(Z0v93`X=@%ueP6ka2nGsjZ=^+1dy$y4S@whOGK8Bm<>_+vzZb9(Ealx<b
zhBZ77YZgI<8y3{(KVHA6lin&?$UO-T-x-A(9;?*R60ew)?7blTY1SQd^h~r5s2I+W
z_sq`bk3^#%Y0y(ZVO@sFOFc)}S<p7kDuW`F5&@gM{lRHU8%x`^Y=W(cE^stB@x4GX
zL6ZV!mi$nGRV>;e4YOGn^3z`Wsx!T8GHH@aN*@!Ra)tYP(%7V_9Wv0sKL_04MxXwZ
zz5MaTJ?!mUbfh5tUC~d}6IN4IMoi2$#ALs{@J5x=-a||<sw=|J{sHS?5L6%P;D^E&
z*2!;nE(3ILN6Q2|Pmr43KN`KzU-eqi40@ix1s#;YOOQ`b*x>UP=P=G=E?z4ibx8S_
z#k+VyoML5e3g3bE#D9m6bZ%+2#NM)|A2WY0v7~X14T?%qOV?yN_!%kC>o1TT_moNP
zW9kCblC{wWf?$I$f!r5muVf?eBJmx*S&6^kZcG3s70wVsVJ8cqHvrJ6YhE#D3t1%0
zo#WK3ik4gx#SGdX-6sm$nG^+DvUUsXHl>ky0meIU*YO4VhIC5kOa#I6Cp4T8-RKv5
z(R@n=y}-|=jeb+7{wBAlrt~gM;)OA^l|o;I7ete(nK6on`!`a(Cm6yXzy;niA-?wn
zrN}s*6i7U8Da4gsuq@}MRJ7z;7ksV#QMS5-{ZlVN@T)7Z-tz$i;+aK<piuB~d5`FP
zEqC)DD4jc05=54>;?AqU8UHlN*t_dbcDYWpuZoK^gRr(P)8CKu42EsTqMv1wTO4^@
z!`cH<r)uHliOq5~1$igdm9z{M6F;HqIzw&V0_~bH54&E<DR=2qy6+bD0v%{hVz>oo
zh5ZilTDYJCpXxn1Y-JwaOkZ@|^?;o9wHXV3qyDfsLy`fbR;>9CDb$slP|bVx^8N7U
zj(qG@0hjcbGy4=;@;`>Ml3y|PPxhK5(Pg}hq-2q)Uq;{HJFUTjgCb&`+;x(IDo9nv
z=N<N36K_jy@g9yPVb;2SPubF3poGMNEuH&xu+60|Oe)PeS{7?S_Wl%w``DYTpZN~d
zUHJ0(*x5#+6?vft2f-Wr2=w(ImT))bE2NaP9T$i47XXuFMMdSDwgj=(AO!w_HfR2G
zLFu0W$0QGxKv9d@A#1N~5Z3uO?TI3QPpb|-OaT-#?Ay&(O7lvFl`P>4mT;6IKKEW>
zJ`Pu0vpwC^PDB(Q1rV>}LfPD&c^6)M*v+k|l?Bnfk84($OoxTA_hrjwmaq`dr#+YB
zc@Nn^ogQxJEnmSpSbOII&2@G(IxJ*3;LuyT9z?hOnxW0<7<=o96ldo{La6d53H+*E
zv}jMH`UWb`rha_eNQrvmsdq^;0D_qr8Hb_GC5WW8cF+3HT6c4p>-?-SxOr5sxT{|Y
z4d$!tZ>ccAq^Am`oDvqndeF$lSlKnXYlcGC8wat%vMqf;gNd;<8EB|{JLEu_g#o@b
zk`4~n!ucojSv(bHw6;36&ToQu`AT3k!q)BslV6;wJ}rmZA7EG?orMa<QsLPPnrI)^
zV(PLNYlun?+C&x50q)}c3aUamy&lUIeF}!C&MVtSn}_^mnx^g0G#jAn#n4Du`37gc
zAS*N1cu=4mb*V=3`L(PB#xxt>s8ha&25^6g`;}XVz&J`+t5DD2A=q<k3}2+`U6+a@
z8}NL!Hie97UO50b+6V}Rn=w`WB0ogsV)iKmv;7RL?3V$l-i}YLd!61pgzM@i2}G86
z%STq0=Hc|aQEI=fPYr@oYCb1U*g^^kF&eP*yqh<wG6$Qe2l!IPy>H&!$EqoCyd3sr
zOc*{{{iGkwl6&%<^@ZJFHoDT_Oybofl+V($(ISJ*oB2BaCY9*#JvBoKH*g4Kkp?Z;
zxcucil|7~(EnAU^pX!q8W*h9n=ry-wRnSh_`#O%Pj4%B0xoX|%ebd(jn$?cU4x1iU
z@iX5SR(V&8H4xILXBY#7<xG`vUc9w?_Aj+YObPa43`uvjGLl`kqnlc#`|}ftb=opZ
z^Ux-mEoI)ye+=OToKe19B40z2vINf1?`G<v7pHdH&>u9L4wA5%s&c=%IoBHu3cDp}
zTv{Vz6!BZOIa_ju?@g)V#@b;n`e=WQuYb%~Xc~qnn*M51Nv91{v9B1nW~TsIjg??=
zUy0P{A}ne|p_kTxrS&qT2K0i%Mz))(%ztX!Z+pJ8^|7e&h#Lc{O7skescbl=`GV8_
z;((-uc}h4)GDS3k(*3>6sk8RMS@Er5uSeaMS|e|GX2gEl(iiftw0yo5a<AvP1g*L(
z!XyrCH=@jy5`ud4N*QySB;|=BY~DnbBR3C{#tqFa7p&?P&yI0>$^e#uxQc``@)qp&
zhV}Zzlx|NC>s~dg2R0~XrML1G|JJu!;=lT*9%qG~NT{*q{z)y1TFB2~TfNH9YTa2T
zxCk@@a>;(uk9|+w!^5~{krD`!CZ4AOT5jeViDJKS_R-Go$m=@36L6L?Zj4x*8U1KA
zMrhQY61#59Ms%S!DD)jwH^TVXimsmGBIs;{f2BD%f7i$U3rrg7{1+RPXGLb^rMTO)
zunn^Rad3N`$lDgI^nsT{R=j?;t~3bleDhF2Is!CkDaV@1H4rq*f9_`AbaP_2F@_<S
z-8r`);^9i>r0w9a%U+I}U^glgy!Vd!t{|SMRU;SguaG%3#VxtwSE?ns(s@<g157=n
zwjmpV27Lv7OW49;K*_?Xrr~sTe#@;1RJ(jEYH2mM{+329*fuJ$$f+2L=+^VO>$kd%
zPeHG*F~HtK2x=2&Cpki=toc%O7*9(}%|NJ>mo-;zbAtVf&As9n5JA8vzbIH<02ime
zhXr?53jdENPbY7%E9uquyYoHwHWn_UZi3V<^j0!>)qxLBwXb-Zi_qu%;NeXfgfjkI
z26qZjILTOgTFB)VmX6K8pb3A>Ax&yNb(0tA5bno)9%zY6)Rfx!x)fSmr8L*5ts2Op
zt*uZwVb{5FU4_<|NTFtoNzpq)*cBEal`a_x6@#f9NBzhOloHF%$C8N>;k;d2irDJS
zk7=OIp-CS&>PM|%@K7x0WWV|nnY#O!IU1);yY<E5+eW2Q5aOmNK~)7y=rqA)&EA)s
z2CxDT>8Z>vGG8kVXgFA8+hUn|LD?}S3M}F{?%8-wpaHN1;bFHO97MuTo{JCyZc3WU
zjVP#1dCPcvX5kzLue-)D*Gx?g-UfGXCS)JuVhhhM^?Ncl9L;1eqFBH_WVo5pugvOk
z-e#%v1|m5-D|Ju3XOzmw!0h;aayt}4#mNbK5E~Cx^mzXWp1RA6kjOT5u4L4DiBU%!
z&2~?oZoNgIzCNdy3k!BU8HlxcMg%1x+@WRa)JVd(?)~tzZ5kc0AR&v9q}fN0M{zOi
zTMhmT{P}&!bnKxv_(UlGg5LHVQZ59PMI^@Wy}jIEZ&QKx_Bid`9-Bi83!0{87JjQ*
zU2<9T3V0(*(#wO|O1PyRUg4bZ>Ht)yJ=hb+RiHnVf1qH-$dM;uL*Ilz63dIk9MNwj
zyEBtUgOHC{#^u-@$R+7ou0GwwR=ld<J@p#6+19P@i9tWy4lh2Mh8sj5bkVn@Ad@;J
ze~B1$=<$SFuJu~}Ozf}reV}zvy8VIhTUHRg+>u+19H(h!eib8q3OjPam4b^Mj%vug
z700h&OSgS&!(#kP^kFKn-MLYLNKd)G^@k>CmX<gyydc!{hP}Lt)1E;rq`}otQvQSK
zZBtov`qpI&Ns3qM7UCKcBVL-jI<-f5CI5z*{)1_v>v=&IR!tap!qB+!-^h#Nx+tZ}
z6j}8HveooUlZqEkwB6;81}4F7khn6xLYvNO@^|i|ZRG4q5H{autif5)+LMZS9l@E{
zWnw%BaREZCcx>kDB+YBl{7I+6cA!fVd~Z+VJue*N@eV#<gwbk-2E-nwL7N4WaqhM_
zfcMz04o7fd^alCcBYv2h1fNR~-wA;d*N;msXw^}?S9CYkX)h&H6O{C&d`gP}b+g$l
zH{Os_FjB`%Ir1c614Q7}On(wn{Czq*+%WWN*__fUxSJ6W)6_5}cSADz8_?W3IhyyI
z0fFZ($^6=!BOQpCktt?Q`F9m^l?xh6i<z#g7w`N4pYD!a-vz`x@C6HXRCG}t^__)e
z;jKw>6VecVzmNik(*0%Naw<f6{!mcMY8(6w+<4IWZWhq{;R`gg-3ysVl_f(!Ioh|9
zz<><Y<Z#3~O1pt|`)4?WifLF2j;qZqwd!DfAgIban!sbYfC!qI&>hIqqQG+dYP>WA
zPrA06*puH|1%AbvJXO|Rin_b!SjKcMYY-Rq1$%S31Sqil3z3<*QgaIJs2Zr%w;nVs
zngr;Az1`rWVS!`ki^97G=u3Eo&L04*ggzUUE442^#&DE=+nB0tI0THRvUDeg6lrcO
z@I$rVMNL2%DK`bOBM^hheY96J{PoVoF*gbyn`W#ADZamDna}h4W=joBb4sJNrmjj!
zGF1z5)-?O>T#ydag>)7-77qtKaNwFdP_Yy0a0&!5aSB4E0*-&Hfd_qfb~7YRqf3p|
ztYoAV(4{x<P+>^(aBVi#b{K+rweJIW<f(lp-}QbQZ|jVHF^gBweuO+>Fi+f}2NNN(
z4At<bp}-NFL^c}&v8$~|TJp_y9`AK|lS=5OKp>BNBG%Z2^SUxlU;s!<3jbi@Dxdci
zA64&Ui`h);`s*7Uj#4{Nze~k`a|)4w>L#yFY1B*HkU_v0j)l`k#i8!S+k{!F6u^?!
zM>Lozc)w&+V?gEAM#&5oiGD&`utUoklnDyK9st*N+b|QFe1{a!Dz^>xP@{9Cu3v!m
zyAO9kG6q?!DF4WU%CX6fdD69~Edqhb8ELd*)WK`KpAdKkE*w@y<D4X1I(zbCS8x;{
zUCFeOzdNGg7J65IhdYuq%~k?-m@ZJB=z$ZzXeBSk@0nfn%6vq}R-6x_`k3_i=w9BG
z07_isAa^ROxanJ{lR@lp9r?(ByExP>$AmZ+ehf14fh=7a{{)sla^0lO#!QhI(}m$A
zyeMT~wQ4m)w!dT(0)<9oxgezC=|7#UO!rX;;69AaKtTXI;4=v1aq0NC>%if&33&u!
zFNvd#tp-S31Y0*q*uw?=k~~=K)T(to5m--6#GX3bogILAW`94S5o5s1;noV)dv}e-
zsxbd5HxbaG(N7TV+A)GtKi|Sqk%17DB@3q9=3NR+2;<^Rh1kzgt<AzP>yyzq%RrB_
z!Ol*hd=TyW1FEx*?CpDAeB(K{584En<SqnpJBUOpW$)m?FW9n$d-o+e65u68+w2%w
zlDA~9q=A3VgESSnQ07konvUr1-uzo_f;o7R3`qD|jL`_gLWSi2JQGV$y(abSd9RA~
z1H3K}sjtFJzu3WR(#Mz|rdJYez7&m}PGkMvPlrEdJS|ioz#uC|DYJl&we4EBq^J$S
zz<}He3bsdie|;=-i3y(w&D>KtU{DgbZJ{dhcUSH3Z(0S!K=fN?UwwHiBG+gANL2ur
zY(K&vS@O5pyCJL1tlN0F>R-dOqt`knGg!Oi<5$0zNlzb62qyyf96(A1?=r|#{|@|G
za9L`&-<ib-vS{M&Z)G9fH=1+u)q<%K-2&RQ%GKZ<LyXm3wj#iONFN0wQkQUH3CGUM
zl9@Hi+>rs`?sMc&5wscCoT!vZEX~qrbrEv=d|n6PJBtrOhS!GWPVYi)k--)M?>Q2A
z8@c~(@-*J_$+U%0K%k_5^_d!A;i<lFIIN#`1aj|EkG^9MN%Nu}FglmmU}7<3i9)wp
z&n{}!h29w9F8Z+3z`CB2M&}HEpM<GiuzZaMH+gLei0{c;lLmUSw3(9>v)tU`1s7xX
z6>&R?imy~iqcKyFMus*O-oxhp9<s2FWfSjI)e3ve`kc|CZ>p2G67rV|leSJ9Y<SLk
z%K5U8d#S4S7rG6d3T1G`yfP|%htp0udT_lfeccA;gVMnmeiX19`xj)}$llH5{n=>Y
z+bZ;OLpkwtdn23ooQ&l=EDRJLy89sWmUHQm-$?|@!xm86o^SClY>rbzx~WQ~PUP(!
z(`f!Rz7Ue|h17ofd#|8owU^?Y#y&zq!>q=5(e$;FJOm~ro_55{xHcAd-BDRnRWd?k
zoL`f9t>Rd#@o1Z7q4ysZ3z{Q@5cAL7T8g+}U4tH=w@FF8GjpyWl8}=7*lzx^8<Ddl
zHiZFjce*(&4*blG+}f}Tklot&N6L<-HZ^WlfX-ZuB13Qb+(3>3ud``MIE*3}qqAaV
zmPZnhtqx$VS<d_9PFczu3<)G<UkAq3`GPT`M~Z=C%UI=Gx!BDpk(}Xm9$X=rTU1c{
zz9Fishrq>NZ@+pppkf=>&s{W0JoO$IagwxVY0)_HZB>wsuGkf6aZ+6G3TN$dkS2M%
zJ(0C-IPMRZ3j&WF%Qt420mc<aD@+Sa<fzMQe0p5vl$5eEne464TslDmpFTfX=E$uZ
z^US0XG`1)LcYPGbO|i#4<PJO{zarqeshN+`z6c`?H7j4ra|V$ghB`-!j-_@ukOv%G
z`LgHoDX=Zm;-<moCB$|V>kOP20WP5E<agH4(4b{TZ=-rOAmSbYU|mv!vISNygHs+`
zi0LkpE{7ZwYeb+iBd=vOmqL1?tX!5@?O*j`64L5W_~^>C+9r1jH$Rq9ijEUf535}2
z*DmXForJx>v=sOAAJ=vCf&&mZn&JRUmne^<>hJR6Z6;hwegQ*mlGL+fMQ)$4%McR6
zJSd94PINOyoPc+w!~jyq%V$J<sCZ3<Ly2QZ<r1P5U~0Ele!&CYn8|K-Q&yn(xO-2V
zQeV&{db>#?2g+&*7>@T+TLifn`!xxy&Pa2r7maYn>8kp&BaEklDt@%OMV3Ff;Hk;Q
zpU>aQZIyQ8Z*agj{t#_};>4F$Hlf>&t;Pvrs*pMA6&qy1iq}x|<fi^ej*0LzB+_Y+
zT|-U(zWojbzaQsv2P##CwE9VyWMe&v_E^L|Z!|$pO=GrhXr#^`7S$}J?pZQR3{J80
zM*biwiWvBw%w0;zjF8STrxFIq-RleO5G%e+POc*=2qGVXPnua{=sSj*dtDjwF@QdT
z`LRZIxa>Y*(%Z);!*^ZCfvMc}sa>bzSS`KGegZad%}y2$?N_aQA*vkZ#knI&SBv{0
zwCyg+pO=B5sP}v5mCEJokyNhi2siNxUxaAkW{YjUsOBLDSMo3atTANvZczz{E~nd&
z())T5h4@Oq5nq#Q;sPcPX<Lb4^@!;>_8(P4^YId-JI*tD=Jv5Si2|HbN85_Kx=a7C
zrbE<a(Hs6&u|*4KCN(cwb<VvCl(MEFI?M{6lK&m-+i-Kksda@L)LB1xf;_$(^k+XO
zv?|FI6({~8QE`b8FNAagK@F@4;#854x+k8=NIFY5?%Z6HS29NwpL(V9<;+29G`HWW
zs!I=Cv!4U!frnPM-Gi?1o-KzE@KQ?46T#D3EU{MfUzB@Tkw-9R%W@>kC4JJ9<r*#Q
zbx*%%610X}Ea{WvrN*|FrYg=^gwl>#I;{PsSoUHwWx+BkrzEFnk~hw-;=wMyi8XH4
z4BP3<k!$sV5m*#Eh4h&0i5n&iWn^D{?ua~L=`-+63u?)-tBE%MV+L(_un&O)Jo4ij
zReAG-3&lR{RyihDUcMw1z%>OCC{VYaKKV0s>V3lZ1M_X1ywwAv=iTp>7p6BPv3!~O
zLmZx7FqKBH&waiuW_u{vlWF6(9=!8{bM^w_5v+m@j3Ft81GEPCY@Qx)G(n0|G9#=Z
zybXk!3PFrk6a#<XhH0DEno9%ptH|S217^0u?RpuFTlYTpLngn#*sG#7*PV@XaB;Wb
zu=CLxF*~f{GURcTO>smB*wM*gXN(cb6Hs#UBKVn{X``qjJ%MBezgK0A38_Y&#lJoH
zON8v_FPH!dYDG~sZmz9ZZvIVUh{`(&wG6u$pbrG`gv4G=+{GJ+j#tsLP_D#)#%%9M
z*|5^je<vA7sz=9$on28ybxUmllUNw*D$s8}5Vh3~DL$?KO`n@_a#y>4L^Jz6EUigg
zxdCFU=BnO_;vPL6cMvQ%lewx$qn4`pPjOV^)<8qGvEEd1Tr82!#WZUc;)cUpN>O=!
z{1D$~ufk=%*?pFS`-zZOz9)Z^v~V_x!klWDh6ZcSF_TibMyn#&JRC`HFw$PiMsj$r
z`TZxo!hGSRIB8AiY}pLBCR}d!Y1g5F5FLoQk>IHPjcyOfr&@N^nTO-q`30%D)b8q%
zH8}zk#XZv17D&-Vthv2x$tZ7hM}Pzb^|Ic8TEMy=bfoO9&XEIog!nS<=5<Vk40dZC
z=}m@2lll)xFq5N7ekWYHg+*k+qlM<@W#?K+ycp`v>&t-eAIgxHW=fPEz~f9xJ*HAF
zx1bT=YdYT<a1}KjjM}bBN>^liVjadm))1`tKpug%f)qEV5TRuscTQaovVWRO0m=v6
z=K7>`?E+ckfv&Cdi^T2B@9&$~$AtBC&of|3^f}uo9m&KR=f#3-7taSr$iKtQ(`}kZ
zuQp#vd(Kj}W!XSTJiNqK)io_0JGeLm2&vzdpk1;dL17r8>};~hii_fE$vC<&*dD|H
z!&{WVBu894XS|e7@1B#T2CD4Wd<qfv-4QPNEREY=W|HQPp=zu39@+7(vFd2z$t@Jd
zlv{09jqkE{?q5()R2Kqdc!$2?1v|EQfG1Pdsy<39xpZj6k$xoDJCmB@FIo<(+K*+f
zc3>ZtX&0TlPafnN*?868YE`S;lC305CbQ2^3DPYvEv(>ILBBJPfn7B=96!tgdUZJ?
zhxvvD`{`WSY9qG9oW+O80^QXXR7o}FVa#xQ`z#Xx1P?ww8RAtWYO5Iho{Nl%(`Wvc
ztm&4(%^EVvCcpr<Hk;TP#z?cpYNPve$ED2sPR9#a@~6KKUwWTu^k+@KE<!+lwmLoT
zWt^4L`-)F{<nzOyN}G=Q<#kVQ8UoGCsT=a<%LYyWdC^%t*2p_DB6zo`&Hb4*sOY2m
zFvBP0N3St*fZj9|)hGoOF;$d@EeU-2Cww;M7}K+9S6~><@V2ge1gaP><3eAt1`J8q
z$kp@_fD~fSupa9D*o4=nhi9KhZJFD@{Kw>t@h6lj+0_;2v%ZUDWRMtB0c+e4viIJO
z8Zt}-S%CaO5&SUzy*i4c0!Oc`YZ8kJhe`aW@u>}|a%2*AlZ+}HiYfI7qut0r_|D#7
z{}-Lu*}>I)-8c!Q&VL}Y@mq-C;{+FW`2)83rkuvo^pVJ3#AR4iHyWxPo^~7B#_>%~
zg4!!?S7kPinR^wk9q?PThJsZ3LQSOTno`3n0HJ&@y{YuODnFtCycNXbDXZ{dAH>i7
z#}GCRXDR0(9(d|?`F!yD=mF;}puO&(W-@aZ2(+j<1S^`{YRra6PmxqLG7N!f;}`Nu
z6K>|jJA%JQ1uz%?%}L<R1;F}y-5uM+{&N^jly)LldMzb)rVk1i{cm;(bZH!-c&b$3
zF^`Q8u;eyC&{48TtUbN@-aKVd)F|_jQKT&%GIxC=*F*enx(an9ZB!JT^SZz28?7s}
zbASDGx^THe*BKdD7@@&QsRCGaga?>Zg}Q;xcrMExXp%{afJYi{MVyR5Q`%Y}yQa^s
z-BEy38Nn}dtrlP}qd-U;vu6DE5Cddt4d}19T9`qWJhWH=DJ<b!0M9~JWb>?ldTL`q
zE?a=k`Zmc1c?i8Zi{D4$cG<cbz~5i$tT;xrN(lf!JS);tv_mRrg;4%#Tf=<_B%N!9
zdk0KM)Wdr5;L<%PMgo8_xOU%fEi8u9-$2gpBJ^f5=oVOCA^y%VocNdC-0o`r3RVrH
z_3C;KDEOvd@ad^7*7;t_UV%Rh>G5(EPR;r|<6Z!Qr^4t3g4!6z9um*(=l>!NjF`&E
z4d0LGpB*Ghes08SHzADLt?Ff*)yKAciiyDNsN%O*Yql9Ix5=Gg1X5D~foLhxxjjdi
z-O;g`DQWvY@I5%y*J;&SOR#-*n3nuI9B%UEL-=SaN|blz1l=}h0Nm<;8uAijO5wsT
z_^HI6v8(mOpVw_q<GW-H?Ma54&l7>TX>dIH4uatIykB|{k!40%QDKx<OE)r}OJX?J
zS$X<+BDhZ#A2r`}m>dz=s$-zO-2<_Fgy6~-%pXkIa7nls(?Yc&z=2wGl-s*{jJgqt
z4(^uS&bEs+Gtd}7$F4n5u{9|wQEF#E*?bZeQo<k{S~X$_vsAvxbQhKwA!xp|$@0+p
z_PeT4RcpMW$AW!iPAG=r9_tC&f32eb9LU0iT5+(@dZl9Dhi9~&Q$L$|9(g`#Coru&
z{{Z*!yxOmN5f`cyS%HCg7|ekF#d_)=i#A?EJf-Nnxlx(-%jaT!N?(#!P%-lTgDpPq
zRxHAZYk7yv*XLjUzHRSpIm3cbZfR3UbZ!4hC=NRbkcLK>$$zS)G$(O(yiWqF9ZAbi
z@1cnm%&PWTE<a%PXr&}6#G{`&|NX;i*oqc}vOu6Woq?$#M(b~#`!2X8?hsz_W|lCz
zY<Lxy$JW7p<EAPXFMtSd#pNXYHGg2fi_9UfOa|#r<-&!0;qAd97|lF8;PWEMl&3$l
z&<SCfeT7`Wb?EQi#1Velg<`<2FCY0@M~+~xgXsox8|L|`_eqn{=S1ydYDIOsL%P=d
zLFyc`|M0ljHMKEd6yisQI&04p!eedX@s``LQbTTqm>SxwI<w|vIr{Tz@O(d2T(hj|
zTF)Zg45H6wxB6Qq2X~D<o`#D1X02eFi*hy-SrQv~wjr+u6Fxn%<Gvdxbr!|9CIq-F
zaaUVV4Jr&6J#Q+zyY@)xADMm|O2x%8zL+V4{WFS(=emyP-yo#%I}QcyMjMnn=&f}~
zdf^)%(xU(H`37r**@Dp)sc!ON#z>sPjMasSv#hCD;`Zkn-QNvke3YLnm3!_(nP*XH
zP2V39cP`(uh)I_0T5y5*51axVVukbP7?!)^b6fTgEbJ{<WE71=FF`k5Ysu5W%XF4Q
zr>bFr*qJ&H_I`^h?RO??CW(MK)KH7{xzd^eGDlB$r&RjOD~N^4^$GtxUB6Ix7hQN<
z5z=$EFII6qbR;xVLg!-b09nwiJT(yxmGqdDQghKDo?f3j+A_yj9p^89qPm?YpdrTP
z5Bff%b{2O6rA^&tT&dPR+>V|U*<C*YV3BYsqg+mBvpyULHtCT!kEcvL*WV@ipZx=y
z73=vHh7nn`Ek1gj*^(1JJ1pE*U)nFKfC?IVr$pk|sPRDv`o&IF>iui;w+D;sl)l*0
z0t`13Lz5ycs3H8KGA<he=%_&ee@3^Auq6Kx%B&xZHKzHhZ8p<4M_plRRGQGsK8?=X
zX8u9M8c6Ynf%`5Nnz~oXG&o7$;~``qD4y?1(nVF0O22h)t}v{b6jdqWTtlk^HB4NM
zzhh`l=9URB&4QH}a6;6;v@Y`H@@xiF*J4^ihu2F0fL2;ahkqQYE()o_K+dTQ?pE3w
zS=i@h+9KHkRh}`x<!p_T#aNLlr=->V+cJBQiA(E)mlF{`l<irx^r&r~Z)}Y#27!&!
zxMk`bTkGo5*3@gE_3!S3Q!57DT^iYVB#%WHv9e_gxVUe2ZtuczN(?x107H8zAvZ`b
zrQ%PO7`NK&$t2Ve0*9#7cnJmKY+_pag_$M7q6$1|faf<8NG!H{TQ7pyj(R-aok?C{
zp0iC(1N(+Dxicf3_NOW<=@AN0U_V>-4cgc+4JFm)v?3=WKhJAl69lf*ex&W#m<x*I
z;A!$9=p-yj976qqpURd4FFT#Sj<HeZq-w4|gj*~0imEXGdgm`?xE;Wg;RPq+1F5Ws
z6L|9r%a6YCyy4_xSUOn}r12$=zec$;-BlwMDn#uIhHbS-I|R&;*@=iB?#aDZ^&31X
zM~}8#7~z3lc(w|1M&z$uzlwM*2dd2UIHEdXtxcg&<mr20!z$Jd#ogRaJVNHkVHF$L
zG8H0y8Wz6%BT{X<b%MSL6ZC$|wM3q``DycU=R`%ho|I|G?&1%ItE?+a?6FZa@3Y`5
zc%|U-k0hywQHCW53QT!Lo%6nzZMX_o6P@K@^yh|zqu?%YCQM%29ND04+;%h=S-4^7
z=%V4$2H<*=SnP9>+bcbftJ1EX$L15l7DKyX*9kYG8v5!SyO3E#m$n6B(BG{f*K7T{
z+WfAT-vCyVJ@@@JY+C}9_1`0g?$Vq@?`n<EW2^>F;gD+vfWv~sXdo-px4aCBAcA;3
zh$o|${hdrL)V8uC^)ks1d1}3^9{M2L-|iO}j-V${)KDrcNWZq34LC=l>1g3cwLXN2
zPHmC#gfqrM?QHYX+?`!7u;(=e883&k(YDDFRA8a4L@t-GNOQj=&dKzL&>moHYC3&6
zw)fvJeHVPL@V_+p4JxGwFP}m?r64R{59MDMzpoUb2*JTIG2md}*!8bAU?N3~O$X0P
z2x{_iO`Mo(*{XMKV(x_f@v&5;aruo<55n_Nqseh)ysa;IF7ElBDzZo?IN*ca{3<Rk
z4q^-<?0cj52XynM#JPG!c_{N|nz$&Hg5l-}UO8g~x}S3`Ss&F9V3%XC3t&uO#+p8$
z^`wi)EDeAffZ~&{&Ftc$Ht)q*FwwoBsW#Q}T`2AlXRUC>t2I>OFXc&$FB-KGe;_B!
zf=bfwgCfbr0dk?+#;9z(+=PfUJ6=zJR?NRg=qx}{BxOsgk!-In?MFkX_g~VQw^I(8
z!|37AVoI;A!7$(5Gb(_J!1D$w6=4#;xkik&5EdIN<kFTZelTD>vg%b@3Cj7I!^%uj
zmJS;rAg$(HuHKi*qD_6O7WokGST%mLtAz=GdFzs$7pSf)PC`mAhMMIa><xW?>y@pZ
zmTs<-H<3B@T|xDtYmUK!jBMM&3UcPORASzlH5LzuD}ZyVP$deliq0n3%xGLyx}(oq
zLK!3M;Lk?Ej~~ba8-NBtH_v6lySEc|N}N|zsM)?5wq**KC8qQ)9V&$`8?<@d56GLT
zi5~p+92Vc`?aV_wh>88m{%IGjc&LP$unXUVS4PQLL#hmGk#`XrgGiq>(s5%Sy3C}0
zPmL^uuc>MOEX$E%fz(pRCRLPE96n6(<L+l#L%&a)LbyNmg>MckJqNBOQl~xi{(<Uu
z0-I<Gl9)^OC8cbm#pWIubr9mznk2ml{CS9Wy64DNJ88o!qUpEwhjY>$yulJ#dIHCJ
zHMeJS-$yzJAb9Ee9Vm-Tvej^_Y9&POGO<-L@!zcY=!nTRp%WUB*PiB>_}goKQN#Z^
zexXl8@}bq4pC&CZBeAgoC)ac?hdIk)e2W=o2qX??mF^)osj-%Nx`|?CthObstS$Mq
zt(*DAJ(pmFf}S*t6`=y};ME517yl$Xp4@LfUk?;1ai9@`uR;K3(k@A7Bwq)&Z*SI{
z*cUkK#e8aXHyWW!!1K~($3T>x0*0kL$Wh6?Z}`)#Y3`CgL4Hi;M4IBSV3)d6riGz6
ztyqLBM~sHY><{ipiwUn4TE~75Mli9Kbvj%?TcnAU0UklsMZnUznGSsup*+EejkK0T
zogSNpd|dm>zEu+?b20oc<TJE88qnruCE4hm2l}Gf`4#lE_@ao7-R%^S?uwQqeAqcM
zKXy!kddzvhFx`<aH*o;6E-k-qsDtf@2>@jvF1HD5*xsS=1~>=7vBf3HViW}tEv&yG
z(S&4D6m4QW|K-y8&X(GSZEHND!Njx)Z&4n<`2#Jocp6!V&S%f+XB+NoQZNDh?=2(G
z=uL?ivFX+F-{q(24*l~iA|M78Tr`{#40TbJLwY-8T8S0nuBehtvN9eQI<9q1Rn>&5
zdktfO03u61XI#~cu|#gF$Uj9mW}Epnw8a}i3v2;ljnu41ll0y^b*+PlW{s4W?|OK}
z6-%7&!@aT;vCP)m4dxwqns+#%d^T(mI?f}1PtelCNj?>1a5-OAGGr+&ur$br_%}D~
z`998&70Z%2v|=J)BGSWB^q0sVE=<LraQ6&b4^B5$41#3Aa?LDsX(JYSnwhx>cVM*k
zhC_*Y`mV7a9*6FZ;?n;HAKdeOwoKybLft8+;Dv&5?kWBUd+!u$TNJHp?q%DyZQHhO
z+qP|MFYaaAwr$(?s&(q#o807_q)t*VmCBtDV<uz1jEDa9+5UgD-hO?WPYqKcM8w^N
zGaG%R6R(QUmy082P7gyehP;`e$brt{Si>w1H4$wAmg(W`?_^L8CTR2@fof~lDK=uZ
z1>fwtOA5m<t827u({4=oFp3246*9nYzvdh79x(u<_q)-)rYpgY@O(R^+!z;$8)Q|<
zC3rmRD|0^4rsj_#P;=0Rp}TmR^|^`pw`x4Z%Gki2=42ygu*-mNfUR)N*EA1GQxhw+
zE?M(oMBUO4Q0t2%K2(cLZ#L`__@;xclR&0-LjUswQ-eL4V`onISKUCs5lxKGRLP09
zTAxL0pbrulVMz9R<e)^pv$dgjxu?7nVh?ZufnH$g+0rT}Shv!PO+-3WmNWm|Xy3C^
zT|?d<WSLFknmOJ#4mTQG@`t;cfeKh<Gh2Ds$GvFOxba1n2Qv*nIuMZXDA#H<sKDPo
zI;F*oZjdaRr2-U^hj`dAVFi)5VSdsZ@vyUzzh~t`Sxnr^IBa;y;?1aGEW3(`kuIp)
z4pj~MfAd*v#Nh|uY!zkJxbwg3+(vVA#v&)*r)W42O}&R}Cj@nH`RJ`H>fFV;lC-AY
z6orX3#k?6#8>qz(Su}cLh^;4%Dv-+1rst7$3II(=4HzCaR<D6=nS4%G=Hc&e4Wrzq
z*MWB-l_#xP$Bn5ej6%L~AjA5XF{TnwdUn2q=J6cXgVu|CUOf(FTqg*(B&CAFj<GZe
zpN?O3`<Wj-1m8Xr6D==R+ph6n?f=;>m>D-JOYo<UU1pX=nn@+{!1n>se=GoH9qx88
zjMLpUBxYYS9!T+Pv?;^a$P8<5Q)>ncJ&yVXM=F#6caizcKPIGGiuGf^o$cG#*!A<J
zj*4f()efK<PJL`!w%cW}-zRVlt&2op6)iisV7&OeEDU!}C6WEr(;24{OWbs<KsH%=
zUDzGS88&WAv%s=M5Gfyi>9pB8O!fcG!5|%{IzsG8&zpPR`us#H-W7`2^~teKi5Qj9
z)0)SisG7LDm~Ik-EOLh%f(^SAzA-AJsA}6N$56fkKJ24<eq0TGy&+J*1Uu?L*q5BU
ziH5NnjfATf8RGIF9V)jnQzV_EMSK!M4Jhz7V(E5BnCct-XJP=C;SdAp{|1R)UoTCZ
zq8o(Pe-QggI-S}@^y%puoCo>Gs=4xbBcXg$jsQR@i?Q`fP?dMey)`=jPx;>VTX#L*
zNSG|(V{VXpKwJGmNE}7%j2^=PN;u~uChH|#3nDu-g+X&Ft0SrST0PLt5lP!B_4S<v
zpv@wRZ*GcToT-H@JAd3YdtOETc9}Bkp-*5Ju<nPNHaHC8fqpf-Fi@!cT76DR#;~G_
zP)V35;X5xi*(fjgwl3T*^AAJg-y^<ckb70;6ZK{L+EDg)(M#spXR>^aI^UxSTI8pf
zX;A@8-jlTX5sc?<uw_Zf%a5)k0^pjq0M6CDA*4<NS;6CA9liFyAuMrlNq)qMc#d_S
zG+iZf(ag#3CN+8FKXJ>Y7V^(5Z>~h-<s1P<f2!6x^im2Fmbpv_T~Fd<AmgszifYoC
z*(IA)0h(Z3$;)Z2LWQH}3>-bj>lX~uvazG8trr$j;n;5SKI(F_>0uHu5;E?4dq;#O
zFEaOtAhzX5()nHuqK>$qgnSmPV=4Ca6#Wf$u*NZjEJW~bL736EJ}M-iHKB*YKA9NV
zE-d82#+YwPId;(f^8lF8i`XnvMbV`A6*7!(+>nlP{@E6@*O~HD$05(B^|J76G?J}J
zk!?U;F2vPdFN(GX?1MU-K^aszT)Q`Y?UQ|Is*ll(5%!^K7VBLP<7a`$?-g(zntZoC
zY(K?A7bvifmOp;>X0-ZTHEdA7b#HpPDo1NxmFUc?z(OF%q1~1TL#+HOz%UX<(*B6t
zJt3xb)o}jCzCN;J`t(qc#UJ+|{yQ0TjO8)Rp*)<OoKlH(){k%Jfs0-TcXw<=cZ<dJ
zT&t=Dn#tnD2%s1?hLT|?C3armpQti$zZo2zL<@eEvhT6^UHsCH3g<v6Y0XwWj|`%f
zx!Q>tzOpomDHLJ4Dna&B<f=6^=#*4zgAnjQ4d)I8sm9`}jZ?wDGrV;~z1<Q%kfZI$
zIAxK&BS!CF0<3`=vShBY!PYjFd*r5?%bg2}0Mct?f)(41Q_eXvsf2{&+mh1A-l8lD
zy{}5kqlR<YQK!xg)lxf%LnR}q0*Fz_-YL&T3B~9nu~W!!9IjG9zDY?N6o!ndJd>Xs
zNdip_Q-9i-*>_z9lH?ez1X{arlK4{IMq>DB0S4MTTWRaSGE>iaOI`qboq2*>0bnoA
zO8r=%{M#b_uW+axmwZ@vNWTsT9A&$`@ECl`()YDv4wsNBd1JGR{QM>O?ZICtp!vSz
zws}?uj3*X}rE<DQ!yRw6{RFBVK&igxa=I2wIPz%O@#Yk*1d$fJ4N)at8I*}F$ghg!
z>Lurr-`_glS8!C^jSM{IE{RiEU38c(&Jc2pU)z*vrR`*n%uK~IWV{9+(W9Z+P6wMS
za9>)oeEDA?t`WKxkwO?sOdR&4kZidUwH4SX7@_(y!^@ImeZr=Awd@t9SO?!jCY)$F
zOaAUYNPsdEVEb*5R4OW!ft2brCC9Kbnx}-80Nt`zlS(lh9mnM3FirkOjQ*gzN>jJm
zWn^Kkf@A$AEUH#E?WGG7A5y71UF^+-f^GO40$WFicI(-im-?|j!>B`)IeqTe;5!`y
z1v9XtusH=ZmrGhsbWfX*WG;+<`9s{wdN%!24m_ZDm5rl#{OK#VpQTQ>vf=4Ki<a`{
z-%`lUJJB-Azk6&Or)^3k&hYZW73J{LtP1fnonw+nsz^%)aP0g<1BDDfp+Dg!5LDa;
zRO-ja#c+a($!m}DQMR3zQJ=tR-je$%rOvWld&vfEYrrWUjr5LF9k<3vlB7T1PECbl
z>E;|+Dq!HN%R$>5YKlS>mAg~MaD+EPTk=^kE-;fBzRV4k*Mfn)SeqZ1q}Z;k!g0=~
z_zEyW`083~n<F}q)@+^Ybp|zh9-UK^`2bS(&qP$eBd&+%J8y7(u9cve?vI!Xbat>s
zV7lgG(dHt?5ZYXFI#95^y`@S2&WiE7D_jwoR9$?rGp{4TwPUXi&i-b;_Fd#4ni5Xm
z+JhZzhC|HuFK*|oYgU<-S3sOD8<f1@-fq=hfz)?TM0DZU1v$6cCa!_;4l@!?ox9N0
zMVy$Qj++a743h957U5AV0AlE;To6|1|09q{0cmQVA^WsGVaD#8-(nd6#T45Hfk9_0
zU+fEPiZ{16PFxx!B3C!?Hwtoq!E8r;vnI3Wh(vR_6LQbHT&R&Wn$l-rAxUgBd18IT
z@4==;c~Vvsm0Q`37&_!ZFwuNDf>_Rkjz@@bXS()xQ_W<HzXbVW!p7lK@VlBopkKS6
z7;LF^qkYe6ZwHyy4DP^!$hmgZl^ZvpuR0W3i`bmd;Pn>$Q#L%LbZD@P5$IPff~yAT
z<`EnmMAyQsOW^<4l5@JAV$Y%VF-}g*3n~|(`BXN?z54A!h+yK;dHB>BQ<%Q5!pcw2
znhYhAP6B-gw~c}0u39<rJHgcLT(RzSN%BP(roIc2+*7OgLa@IGL`)x0JCdlZQGn5F
zW(pU1SUnrYt)w!Cz+SQV?+)Y{XCQhu<aw(vf1nzL`9eX@5nXw0^cKfq(*uSexd{Lg
zLPLt4%OYzSbnm4Ogn1yXXdWu62}8-tHx;ul5=z3<p+>^K*!^*%LXzwd&8@3y$via<
z$MIUa9CW}GXo8GFaz7&*#4ZSOBWK1(FVQ1AIgjGmW1h2P)$g>z?P$0iqAw^?zsu*v
zF`4#?xW%kwlh$Y}K(4IJp6Td9(Z;O>vwxilSxaGXYoE@r$UQ{Rprkl*yG-+CdjJYn
zKpGU(Z|Tt(GSpjt(D2>=IR?Nn7@gM1>N|43#GlgFO^+Di%M}xla>h$j?qu*9Tg*Ck
zG$Wxe1woEZx)EPkA0ysn9}aC^1t;+hF(t(3#5m4FJpS1YT;0iliB04<TUU65_f)I!
zfcC&z*=}W~TxBROh^bhe1p*Lrd!>Zt^eb$<3Bvwb2Qa{<gO+9XG;$M&@c9T@IvgDy
z?;4|{)jt<s2aN`*iyfb(P1K&*$jQ`NqljlAB^#B+j>oarI8+hErg^(7@a4=~-gF49
zR4|U+1^Y-LUA!YP{iX>bNLjjLHrwfjxWpxQ1zJyaRkz*jdD(~jhiHr;^PUfXckMI>
zRtNh;i2sYtpPm~gD^7R*GQL4uMo|jhJ)RJRuQthBpA>&I`TZj6OwQS1JDe!Ju#oO2
z#%K^Ck5gIsQjQ|7rnzD>M)0p7(#PEww$0L>ZqYK*{@aJZ7hK77a#1VwbJd94Y!LQ=
zEWphAi04f5ZU~1+i7*|w6rQJE7p_{2JAZvq$-4-&QZyyk=s?KC{=-s#6Z$^Jg&Tf~
z8FdC2Rbl-YGc0f1$;`10#E(0&c<!TRk>MZ3lGES)edvT`mM?BRxMrx(Wcx$uRo+(u
zf@%P}nA(K+GxTs^|8;L2c!!-k0(`tJ>Q1Ln5LC_($?JI35$;06B8h~)pEqEEvibJQ
zA8ChSlUQIHX<G)zPYIZgib=X4>6Hg63J|IdlF*aR!`bia2@Ryyb>&@UQwa)W3|3g7
zO&9HNSH~j1jQ-^i#?@+kFXH}B+aTna?Oy#2hydq`MaTg|Qr}S}^u_n%w0f40eo_IV
zatVyRzquF3x{nlQ_O7D=5vjGWEOZD)Ny4hp16?pG{QNZfh&qC56`jEj*SC_hxl+pK
z7CW|zTi{uqHjPz=YCM(8bF6#ObuRQgnqqczi1)8A)C$l_O+Z!OS@+`s`px9Q=6kkl
z7?PR(fM%xss7v+XP~un@)<!r@JOo+usp`vyAk9Vf-4X?lm%Tar?c<m^+o8GCVM|%R
zmyt}21==A~3I$N#Sn0*-AoOt4|6W`O3B5@^eiZ*_jT=y57}&=+6xMr<$VyGV%T|HJ
zpr$BRklWTIn7zgVtL2@_8;atY`0NcT1uqY`jYTL-i<LVlHk5aFL==f(OlcMV8Yei9
ztWG!t536xkV5+#%2P1H&lCUF4ww?ke&fBOY{^#R+D^N_Ses1Q>*dxqDpxL^P8)j;2
zd>JHZ&(g2s)Jpr9Kdtd5lS|8Kk@`2MibCAvQS_BZyr{8|V^KAjcu(OLvMXR3ae5cU
zf`n_8?peLs+=Z{WT~1*)Y<Wcf8*<v=?AFaiQw?6oCn?t=&1}V<RZ6=jaWf^y|E7U=
zyKb`6L1KR!oW=wFQLr347u6`LJtDKFF~p>P*2}&^vVY6&g{RzpkEI~WnaOIxX=B=i
z4Ka(`4|4w8Ko|uBb++!de{nVr1U|6FasbeLIODC}(=E;#w}h34@fbN!ma{lPj)%uO
z(l(n;6RrM%wch*6;!K85H#Q}w9+XNO@2_|aQF-+0u?_cP_?SWke&r6C0MGyxg{Tvs
zbG^h~EvqECoSCE8ydgMHr>3)zT2UHy`?|0w;Y>Im@ZURvcS!tEkem}B9KXgpx}PU1
z&Un=a9;}A9lYW@`ey4p$%7^c-4N>=c8M=X9V0QW3MY9<x-Tdoq88>=cFD`<hgvUI2
zZY(Y<Zf<LM)8)P2vl9{AZ4&Cy+{pzDe+!duAwuu6!M*nKy&8@mWPa0>qDhu}LhxTs
z8>Fes=Qy_fOy1x2x7l`13~7k8*X!^sK~bxDkx`3m;+3PclR>Gt1M=W|7-H9~M{dTI
z*prWJ)?nFez+ZgwP0j>&kLs!W+G;uznUWFr%_I4ISw^}sCilxOn|8nW?7ru2aC$V8
z8iYHNM@;~UHuWM0_O~u!HO49$N=gTF*diA<C%B#wKe_C}Jv)*J4G`{P8qeP3|8OAV
zEzx7**p)#?(NGvQypGgmq($#w0mF74N85|j`G;Rm1FXsi!@wFFS)Y@lDU!gsYqnO^
zoWG<x%dytJO4ZJ;DHx!4s*wYRhwHk;IdSa~773~g4E)fk;g{`fs7AoKB8F;pWn!3a
zmb*orck4nCZmlQ)u5@=GY#fi9uU99KObx)YdAW-u-ZZmVAN<7XkJ5=3V8f4%S4X=a
zh!+Z%;Xz+vH2qGMB{%D1N;)J<jdOzYCg_N4hfbF7%aIu9jVfBs8tb+G*i)PVY5ZMx
z^H+s!ppVci0$;9_$^svJXhJGHIi&1))k0j_f8q4p&l~eK0>ja2(=4;~ekiUHY*IID
zG?D^~{~U6|Wm|!)`%-?wP-iR0df-@!sdlB4CcR9p#EaSYT>JqIOs@z|Im=&v{zN42
z{b5$$B2;Z=7-eUURe{<jq7;earPO0TDg5UX;#WU@Y}K(v$DSi1)b=hm9%gnr?n&<5
zAwFbqK38KxD@b_vMLJ|`W4Z0KTbP6cK-E8Bt#y&Iti6z-;;KwvI>wYh4^YRg^|dz_
zt~-SQ-3cm@W<dm!Wk>dxN=%Osw<5g)YVn#Tre}Qc(x?0K3dcw3eN)$;q|>9hv=RaR
z4*ujC9Ru;;SyIJw^OUet1UUpYS$nVsCHHpug>Tq?$ZO5hfn2||Y-Jz@68Y9nzfjFO
z@Gzt~SS>6bdg${qG?gR-;~uoC820yJ&7v41WW4(1(Rg*z9Ia8hc_af-F7<OPe&%_<
zlbA*RgywEnNXM%BaxsX&IuZQnQHg8d#zhH%r}PHI94U6Nr4c+43`X3$zbD^#q<FL`
z7n!DlL!Jv%Zb%2{A_qC^@sqDWZK5SbMF3c4Xx6d+wI#?EDFmt?Y%VskF32!H$q=4P
zxanRfJ>>erND*so@+30iWm3%Nkj>J>axv;cU2TOCQXUXIBtqNUq@?HERZsKOf7>so
z?Sb~hNXf3;34(|&mcPh)KZ&uH04e%A5En@Clp^Gz0Xs+Ar7(3m)wpD%chKLFu>PIa
zM<V3k+K+}J%133(QuWt4<kA&oE>kIPUSn_|zI11OV*g6sl`V$BKh=&KrvAOek%~CA
zDCRe^?S(VrtGHP$XZO@nHB#9cYsleM;Lf4<`1WlLRF=@Ef9(UdHkeScsnVN75g9LV
zf{;Lmi+{JZ3x>Genc9*AV8+*HOUrN_zL$c*GjUw(p+`7t04ABDFRyG`Q`rYX4zY$1
zXq}{kg&Ln*aBYEI4FDPKq#R{nm$8_2sY@w>x6D;enruc#Bb}36p&j)3g0Avt80$|B
z7Nuupli~Fp--@wdy!r=9o(+eV3Jd};Pvf*LmTuSH_j<Oz8H;%xUZ}dfa1JAFxJeep
z5kcf6{3GMK8IrQqd-pT@n1a<7iGI?X6-_9&@U{!i($dcM5Qrr8Nu&`0M-s7_+I|MV
zJRyH_|AR&k+9t=VNBA-UE~yb;X<BxM`daEf2>CSwEn~nXUZKP?bh`c6j&}jzqJJ+z
z23{kvl1&*DY7UY)4vSkD8*b@KuCi2kftqHII0a*yw;O*5w=sH&E{ru}9??I!D7!OR
z^rzNU4t(oZ>58RPX7$x*5MmW`_;^f7vUbt>XwM>|KncFH?qt}PS=gebT7w26_GpWb
zuf@WlX@hP&xoz8u^2ExU*YQ9<3|in1Lb2L*clS*O5^#goXR|)Zk?ArhBN<r9ByC~D
z3HL>42f=nz+DlUxNaXCG?cyTqL!gv{39~nL#EZ|XMudjwF6FP=brK-CcOyp!BzGw(
zG%OTk;I^|QPz%=H@g(R&iSO$aKomg{J9o|6am-`r#6ve6PYtymI2ru-fQ*I)Mw->V
z?Rt|UF&|UaGSEMyW@tUA?alM8pt5fqW3Csy`2I~YSjgTE-C-clVNzuvEM*~f(N?(I
z(aox|8UEj<`+>7Zo5!dV;C2dbB&DS*)k9&BUK0}P<t967n9=TtxRH6Jh&>HSS%T!n
zUDn=O5NSBv1xFB8nF(F7-(w3Yw_M@1ili7z*VZu6UPU)R?)9^8%roF)fKA@GAQp`n
z0=L$+)gA0by_%!CTP>4Z3QJAJ8hkv30#yN8N*@#<qMq_1RfiCt>PV}Ug8$~1&9cD6
zL}F*C1QIrhjmUy~>!z-_-*WjEt)bvw>S#lMre7EyIsoo~xMm@@s0W3sv20CQ`)Ge)
z$U;HRZF#qOgDUoqv1`dA#L-inhrqV#cq)A3R5luNjnt(}qfz4LTHelYWS(F&l&dlW
z*C%onuDsYwRN}Cz-wa8u8>zZwr|5m#tE%u1a<eodxEWOjU@7(W;S)`W`_flLh)D+t
ze9Dbk8fE$tmBuQ$jLc1l(%rMI%8|#xDjm!4!&ge;RKrcAiDX!jld|<_+}2pLzAzzj
zoCm6smIROekL5d8s_GfHq8a#rFFNtThYF-RSTWxiFyS}}IdE<Z#4V*;26^n{aWEKH
z=rQ_E%5WzZ`>y_$W@&}%dOrHnD9>7zm6DCW^T*2B-}g3}b~XTLXf_1zQKX0=vDK2b
zPE|RbYQO9>Kq;Yjg@GBpEche#3`WtVI0LKP@hIGl97wLJbXEVQ8OIErbR5Sz9}(_t
z$!=)-0^mHd6C0Gbt7-Pt*PBuM8F#4PMwTmx1LYQOg%i%*=#Itt&YK<$tfa0>kngS}
zkSc08ig+X4u1*Gn^?b|1rUZ-vFA+}X1f$-r>bY9Le}Iu57R@9B(ixLk5db(#RTz!=
zL3~>;5jI#zaw9}!L%FnO^8oGtaJf}w*#(2Nq`~@-B3LE@B%4NM<UxomuXJ%!Nz&H!
zi^v3~alwt44*kGbX9hdPx<^nBcNyw?DOW@wHI!aY*ak>2Ypx~P+cV#h%XItNY{=dR
z(H5JOWy~tfKq>bSj4mJp-!?_VKlP+T!~>76#Me%ld{`erp57mAWwT}3{pX*qh`J;k
z%>MK02E0Tx!^9l}!8i)qhZPs%m{6}42Tl3(x2{hT2y<vvCX!_dc$6rjPsb<PD3Or$
z3VN8!gcGEiB}F!#&@BPF#lw)b45T6euu;VYMX)pM)C7@{Bsh|HCSUI|#};Ew^GcHZ
z!8yI<$RlBIJM!H$$GGr)Jh9LH^fkXy2#1xPm5RV=lSd=`Q&6EW6^?NK1_EEi`2^RG
zd2QUwe$VKdjXf%UCWlU$XB3+A)X3MjJVyAyTufmt_J(Xqm(N4+=A|+Aedjbp^lpJV
zSP^%8lFAW-_XPTd84pwccH`gpeyzZ^I|+I8Izm+5SsH4~Ii#7HAQ^caQOqRk>4VD3
zx=+L&E|}EQ&F2V2%+e!yvY2#?qbaknkIWfD@;6AQfejg@1UVcUU2@#eX%N6%KThq+
zAGuHOX^}8~7~Hmz^TOxuT@_!OtF9@CV*+Oycgs`F@o*+Ag~FB+C>9JnjB>8#qn{Q)
z#<%a{vM~qyQ+c#erF6+bC{I>oC7HF(iU~(dj<;$3$49rMsu!miYHX;rjdhZ2)s`=&
zQXsc<FQ8G+e$9M~Je)<F2Eh(=)?f+dsx8eCX9L6hTm(~@VxU%4O2SRdD`T9T_P$Cy
zoWH=1ul|!HXlK_!215Y5_rlb0YlRDZEkj9AyKjqa5*2J5(CH4Va}}MII6iOxggH3Q
zHrR|>&6T0ci8gZ1f|Qn(S@Q8EeHAV-^Mm&OsY1oAgfz0CoHfCGMEqs3RTwA4C{ryW
z5-bt)pNUj5<^P>ZAGQdEorjEGj;GJ$P5)i()U|FFpJ==_EirV&u&&toS+G9aE;rR8
zUuoC7K3w%Tl)<NenESC9?|=(%aB&0#^NKNqL+^8ejTs5Cz&;E!q|15-DV?U(Scf+V
z$$z{Ikwq&QsqE*><tBLL&oiq$yjsgK(nA`{iTj+7eUYLQ>m`6aHW0Oke!Qx+*|hl5
zLP-UpJ;V8RqNM#JtjzeO)ZP?&%DG$GaQj?Tk}cCat?MN=fL;Ryv`0jLu&Mwr_8)kP
zU=PKX7cEC>uOKpp;2^JlR{ujZ+odLKe?nS4bI427#!Xhutx7RXnR=N`T;r(4RDI4W
z;To~wkj}BEa9-s_V4Q&8aVb8$XsUpF;;nK;<Q9O@Oymb<qRe=l-#gdOBb9LoP0}U<
zMhP!5Hfd25w`YZH)Y4{q`*<u#asaeRgT+tQ=veX9AXuYmPY*&Vpp~QDMZh$&3nLSZ
zHkt@lp6(f6g|($4ELvnE;WTdZyu`uij#OD#lpva)Zr-hG5}_$i(wQvF`Rpw`*C>eH
zY0`^%H`xy(eoKL6Yq3nCL-c#FC-qS~gHHzV4{@pdhZRY91bXqw$|rtoUkB@ku2OLN
zV=OGBorPYT)T>N8=Fbcg_4nlWq#7kY(i#DvD`q`0Pid1!a0Df+DnH4PgR@{v9k0OX
z9&_1&t^|9_lbEsf-+BwBZ89#lIPn&CJ67{yl}j$-9dL&1OZX;P?q9a&vE^uKIgff=
z!sWaQ#dbA&n!&uzsI**Pkc`$<?VYxq_|#>fmD+z;YvCGfg(fFfWm-FuO*!>A>1-Bt
z0Y$cpXMtRktx?Cc=MTp2ng1G;9pN+tG8t3W4LSVA;pTr(@1#gGVA(bFKS!sHzqq<p
zcLsI?Q^*Pn4l0cy3%nXF%!0oj%eBMv0jWoTS8Yf`QD^%U+e~O-sS!qhEg*w4*$9D&
zXB*=ms$x9S+jv5p^alr!JNemz5U)hFx4`V>@mbNwtDRp>igWTI<=i?nGH9=gmXJP(
zZ@Ud2Tc*i+rF6$K$&1O5Zv_)MrRpTBTbCo|<ugG>`Abao!J`B9CdWk}#@CrdfOX2~
zBHBeGpj4G&)?6^OLm)%DyP4#{sc+GQ90%MCXz@sb0j^{QC#L#C)QtUsp8FZzDo5Dq
zO2f%K>#g$9D)^}yT=p;#*x}q53t#eEx1()lRef{AVaJH->=U!)lsjYn6lR+Ga+FJb
z5%)-nYJ_eOXfq+Qk0ELv&9_ntLw(_&#t3UW)Fb2WgC>}AZm3|X(o>*E1-y~EGi&fO
zoeTjJ@{tgMSHf!4yqOJOJKiZt>;ky|7_`S6yNmPYg1e@MCw^uNm4)ibwjiABFbTAq
zShhi6!Ytw5M5W7t(7hDPM@Y;g!F|+v-`m%kEI07|Cz$Tx!Uh*QwKtKVo{ahtR;e!0
z$@1!REAT=YV}r_*2k~QziZVf<p*AL*-eQJASTr9+Rc7Y&a$FRaabk!d&irK4i|PX3
z1{$D(Ja6kuZ!0?p_1%Nd8S|`DR||HMze*h6#3;h$zu87sbu{RO6hK=$*AaP#^ZtCJ
zSx!dals6CB_cWtHwOf1JVSVF*(=q9PWCar|!xtzUJw9!b6=$hfKT289bFRRWBYfhd
z()u&zNm!hUG%J*luq;vwJg#)6Re+x8UPZKSeE3b$J7jSXz|&Prbq{RM4yyq9N6)hp
z27>4fltAHR$7>akD5~6GmvX58m=veBf!>{7SnNuawU8RZDIRjc%Qm>tAKe25>{$*D
zUg!(A_mZm`d4w*oZ4gJB{dzB$3auR=-aUH){5|C3`FLsIakAGVwl7lV14V)+-vhsO
zd#bIViJ=msaMvKCF&Pn9qVCvCLl2Me=hg+R%<URk_cRcd-BoVc=#QU-XKME;!p`z#
zbqvh6q-%=({^YrcOkG;SJ9egs5#L27tf0NT+8Cic3$eTsfs~7Mspv(o*n{;*UDmkQ
zSz_BblAy*v#EK#}4C1O0^#Gf5ox=P1%G#2}QhZ4-OYGw{hCrMki<30<a_Qpzi;V~O
z*d%uiOBYNcm5Psrl04{MA6qc}5A7`#V*ZBq53ELnH)LTrh-9)7Ws{!zexWxkGhe2$
zF$<x4<v@c*X_9QVy1BW#X&pCcS^HgX<(VBsbCT57dQgurY;7arxwp<B%j*&t&dd}(
z*$c3SRIwna1jgVlP78y4-xSTScek$kZ|_qh3bEdhh@F38e9;j?p8lT=s|pcp+76AV
zkGOZiTQXn^fj^*QZ$I5_3b-<(2@9BQojgt4r6LX|c-cVCk7gf8Ebeuzwf`vJD+WPR
z&82Q!Q-d!hw*H)>OVBR|$#Fo+unoNHY_G{*RL61hZ>LDF-)%FF@%{KBO7Qx#K?oL3
z?B;mx1Arskz~V=_VD7Y~+it)<s3QgE8r12-%ium{kFnH}PFYJWwn$7}sHKM@{?%0^
zbx0G(n&oCc@&6rdc+BpQ|7k;(BjUvQJi#&qQrG1cX(2O<eZ+&HJaE<mE!FRacg<ci
zJZGFiLo5r=CV8KHw@*4YY<y)aY=>^_NX!?hK1W7=Q;Kw-V=tW~0^dJu7kuJ`CI`+}
z<)=0&dZ&|?t!G?nPqgo~<6_0bXL3T|={NT<CxP0=R~8%-V!sOGboghE_!VtVGw=|*
zhvN8*j!4AT6XaM}ckitLB7&#!MQ|sr+vxInt&Kei)0etON)k4t7gOan1+lEg9KL*t
zZ9CZ#=I?m?v&4d{Sy+7-(Hz(#NURkUV)-|QUO_=4U4I{PQL-dF=ptjhQ9CHm`&&31
zbuLk~$lX$vEIR?zoP;e!|CpVOe`${17~>LmF`8xqw4bE+Fkd|ybVrjzw>S$vwVE>#
z#A{ZfXIBs0lMCbncl24TCnAx<I!xBWY~&p#Kd?4^vMTkJ{R%Xv2`(W|8!DZ0m~wI?
zE?8BMOuz#FkNw2oYRqFp^3`HE5_u|64E*_AqeWt*vEJy|{8#qANb2~?MlhTT=1sFg
zG~iB@t^OqFg3STY!w)ZMWN`xO6AI&1t34+0Krg<wpmxhk%Alzclw3a+PpuB#O@S7x
z7<PDLd<2gD$~D<<*5-_(M{GmLpn#n097IiJ0Olf+qT&Y{-0zZ^i@ua+E!PI5nEg{7
z%i~M%ro|Z(9|4$mP;~7-!xjgp*&ifIJy(B*Y!b63E~ny4Y1VFJkz?q4IdIw^Vt^3+
zj#h><t5bXN{z}{ww5o9tv?fTwR{G81XL?JP!hee;UuxFQ!|o0wQ<SM&Da`wAudH4`
zmih%O$N}$|aS(nP>OYL1M`TaWpfEY%IT^Fx%W7K<Tt#I%|64ED<*rX(o)>^RO9#AY
zxnN|8QT|lF1Rv&B1zAoXINOcHshNoJ7A(QWCY67y5xq9;3?&KA18f+T@jL!X3_p?4
zb<GU{-_vj->1BYob}%6buzYA!)`beEjQbrFY6p~FA*kks`<Sl#@RYg>5ztyDcX?1w
zF0fTsAmHgDC<l8_brj*$Nqvn6E=1eY<XotqAu8x9lEt}RgW#*l@5W-DU*|Pq3TzCW
zIt8IB^S<8c(#4}(nQB+qe)rO%)xszH_`M=2^b=zjs0>QX*v;^OO|)k*nPGs!il7P#
zX1WVH@^RxaQH9FzQ`6@DcSV8HV$`A9u8oSN9CV=gv=R~^?{5|gK<#NhSQT$$aLad6
zB@{u``X(tD0uQY{qQY4{zW~*e(Xl8k!i-gGWe}GS;wP?EL&SA+yPZ+kVO@Z$ke<f)
z;yWGm>GuJUP(6Mn;{pI}Zv(BpR?r-Y5V^UNeFf&qA7{NMwn$`WkWIg&-4UyK8-5t6
z0uy169DKY&C#=x?pU>{@F1R}P{C&o|3W%LBujb2#dg#LGK5G9SUmmfrd(>=JMQ^Z&
zj%|p^%J3=+gseOi4^+DrV>Z2zJ|hTOJh8;2bCnIZ2aypIy>CS#vd}i03*kH?;_hIe
z=8{cYG^SdtA1^d@7T_Reg{Gr<_7k8VC}d+o4|V_=ux-eDv*^mlTTm2_Z|dUFRx(@k
zht&GAQfFaM@Wga$1Cy9d6!+ZFW{;~t%&65ybdvZ<OYb^DP-yG>fH(7xjC#`2sxrvj
z-*G4}Fms`cnHl2VF6yOn;U07cfAw64<f)btubjKdE->R6;6almFsJ(#=S^%*o54Q+
zwNlhObq{(){Snhp%sNRp5gf&3<Oq*nLea1yV4CgDPrzWG;3}x$J05VR#kcQR{Q*E6
z<kwby45dOh|B~!{@|s`v{i7mACG<mMHb!~J@s!s&(XnSL^ahK)uVR>`_yxs};c2M}
zM)<v=<PAot7{qq@HJ$F~$ts1E<8K>`@8(M=3yy7clhxASkpy_r$JM8E0HxD`&QJh|
z1;oaAQl7-+b1bUqwX_o(t}&==K4EIflSOOK!xR;>5%9g4h(6IwJ`%^a)iSsr2!hZn
zFmc;U|DF*f*3)gSBKC;)eQN`y<&~SN{N}N^xacL$C#h&IK-<&s5|byaGFtw(o!dc@
z@!Rm+rtWq%C&iMOS*Rw<J*Rh%JnnqSP>2Pw0nE>2&8kYW&s9Sr@BmExU%j=IymM*U
zL1Nn{^YhEATlT#$0($SNWh)pKNN7BaCOkc*3+Er8am>E5PbvM=No8f6O&LiF?!BnK
z;#_S1zk@N9p_SKo>PHMm9_=!Z-k}OP`Oms88?UuPlVe1|ma>pvt@UCMGf*5IGCxKS
z8IjSqg=8&muR?erpQtPM9Q)Wx$P(wUH9<%6aGWM|!Ogh3AoNNN7MFBBXD=;t0{|h9
ziQ(s*hPXfbMlPY)i4eAjJfXA*j)4o%BHvA`Z^YHA(<6E12HZ`>KC3gg&fF(O5sJ5`
zih(awG_;n@<q=}9XT(HQpvf2&$CAN}vHnM(wD*A1gqu5Z+&S=xpyY<BdQNUe)qNDY
zg5RRXq=U5B6jp+2owaB*zY4WS|D!SqDH@Eg=L<&^f?)vir%-Z)LJ2G{@=dxWW}2*u
zb{Zhps0Qht;S$~{92ruSy0#>F+Ika3_03*Sb}+&#TMmbY=mW-_r8=dJ5M)*n&crNE
zOboe(X=Y4jIqZ*5GWhQc=A~P8L~WQEV~kx(GN=*81CFBGmS%M6g&Db)xl!a^Yujl8
z`2x7OcB`kLP>7v2SNX^IGNeWZV6wt3{pL+K<|p`ls)s{H(WIa;>7{&b-IH|NfFO|x
z9*IV`oAu^X=&Ns}=xV>%W&hd`mDURs%AZoTF?dr<T{o;U7<N>7h*0k(Kpqe`4lm>H
z_rJNkU--~vQ&c~z2VtCcSW@u7P~dMoYL-Ce6N-MH^PuyEMCB%LM+}#7%T`C)4r4TO
zVGKRe5xeq51X1EQ6jso=BheLrmHQ$CN>@7o&?#pl_sf{QCDpX9Xpb#0S<r;@9`-iB
z?j>ya3x^d2x)VMmV|A%7(>!gS>kSjig6QhkYynBDcFD|Zz>*R*Werh30Lx>&-(*iz
zChK8VgIO*8g)`bV$h}bOprZVE95>R!jC>DBhK@gq{KX$v`kT&l`~Wm3V@ihEA_PUS
zlQp4r5QaFA#j0^Um#5*ut84^It4EIaMq+#z@mu3W<hSx-zqAOvvGPlr?C0F7$3`iN
zn<|8_Wa)sBRGEYZDF2h%qz}A6&5lvo{xf|rX%ba1V&wFct>A|6V(${!BK{`yL{1i;
zcu5G8e~dtpJ$g<TaDo{4WUGpNH7S~@FW+-cKGNe-ISjk_ZQ;&BxR2P1KDzl>*>PWS
zs{?k^pjfgt{u)TaozDY*U&!9`{v{gjku;?q)n`Zj$>Mh{9nRs~)W1v;+n`v^GlFkI
zf8YSRe#zL29AMX@bm{9^C;;vGqlBfC({R=KU4dGE|2TOQC~6pyTdbMf0+FE#?Db?o
zt7w~=NMTj6g;|W0*<U42s-j!Y28f<C_OHl0)4RES11@xuhwCf%<|nKBmW!`bVbp3m
ziPDQozMYBVCOIz8ObMz8)Ao-Jb!uFI;f__GI1u4P5k!}rYEp@mDi^Q%3|tpj<Nen?
z!*g+LEEt3{9RPLg{i|<f?I8mTy(Ui5EvJ`BiEN*)(1wf^G7H|nsWq(C;8QPdo@7(U
z+KC*(Mx;+0g4ND3wD3<2%gMP;mp-bXbkS#z`VTHiaFvvHA@D)cUF<jR6<$wj(1V2E
z1EBeu?H~lDo9mR^A}sdsL`s+z1Ct(3fc3mW1Vm<B>&p6b8E}%YqD|lZtGj0q982%y
zu@uT=TSUVXAP1blKVZDvk$6?oz)s$`tIsl;M5!S9)Xb2$TmC?g=a5phB%1|fg3(uD
zE=tj<2F5j*^JJpSU=eJ)&2U>TIY(eO`kEPp)a;aH9)D+hb>czCKQ62WD6Rnm@tO`a
zt{tS+Yr_87!W@#paxx?f`seQoIxCS5<G`th*GwWkcra?TWiU6l>+1RJZ&#2@OPn9u
zNrhNkx3msMwrKo@gI&1%+iTo?ob~U`AlO6C2`B1Yd9+A6KR&X0juYQ1pO>d_eTck5
z_MjcTE(`0n1MLbX^kSVj=DDwILA_@ax!B5wr=yg9DM&tLDP8bMq+y{sk>QQRhQ85|
z3B#hV_-)P1u<<kbh!e@>ZdjKb_ZR=0CcCmVRC&E)d=gKwJ>nc~YM^GYh$ONaTD3#b
z77%sTNx^1OfmUftX&QRbx^gDG3O4-p7<^ei;QE3nN*mhY;8Lrgw6DX|3s&tTcMzFw
zXjWPB{Jr?)2lN1wPQZW4aKeT}i)P;e$U~K3%qP!|Wk3jEMHpsG@=qe`%>$cp^ra#m
zyT^!Nt-}LR4~_-%@q+vEYVc>9CpFcMf|#umw~~#3?&nYQDZ2O@Z2|DQ-v!B$k~byD
z{Hr0?9DoWcjX(+<i&V07&6u@B7%Q3}My7oHZl<_K*7N&@tiN67^r)A~KR_3d@*f?y
zI8yQgP>E5x%?-8MNL3WP60@zL4}9{rX&Wzx1ZvAgL9(5X8p8Hbo0w9ykaUT54XTq-
zy$4M6uGbH}E)7^VhR3?9G5sFB!aU-&l9Rr}P>7^Q$i%;$6jwqsk?3A-$k>e*-}4_c
zlOQnUGA&KxKL1RYshx@$y*L%lUW{<D{*Fh@5H}ar0Mim}#sl~S*Ef!Wb&#XO!Q#Im
zYM>W4=8MNdaqZmgbqb`gK9>N4eCCg^<H?d%asg7za#7&(sN<&Du@5=rWgwXrb_vNp
z%+RfCPz*Tx14Aa6f~l5aU%@*_?6ev?i*bBtIFS)7$I=fvGvWNqC#XBcp*s|9`m7=&
z^~JbR84`I35TfJ~M&V2XZp$TWe7(PS%M)u8H3{U0g~@d)@}&+RT_>Dc!sM}1+y%^;
zoi$~i*cSLYm#yXh7s(53|0#I^pN*LT|9^ZS=|nB8olP9^=|rs!oK1vHjO>g}AbEKq
zotzy_3~V6X*KJf}?DpxAJ5SWy;ZoLthFryeU*FX~5IC&bsFW}15-@Cu$LGTrUG?b2
z=>?|Qz~$dPym}1dPHwfNO$n=9O-uKSP3umzbvYMj6z#N>?q1VcO-#%&fBjB^RElXh
zA$y*0IvRew<kc~|D7nU?_Si}elQI!qULm&Gz8F@jh3oKR>LERHcoDlDyH><|Z_lug
z-#P!*m(ZJL(lTxlIY#|r-fFYWA|u7+IBAY&Q(4C~IKY+$qp;?T+*pxWg7I91JqIMv
zP)YlucicIe%LPNV<DKr}sLt}iqE54Zw<MeKdD&y;1ztF*X4Kj0Z1~W!aJ6*e$bkiA
zM2AzQQqAbPGXMHn#f8c8J-ZwC_+ZB%-BORQuCJPWYow8BvDucXJ@g(*6{&-iR{CN~
zn~mJ^Eh0@FKY4{zX@pE=Y*Zg6w~w*`NGd4l-$d>2-w%gG1)ES;&n8!^WmhLnmsVun
z2joHaEE{q`tFTgzpA-{0y0Kw1KiLDEV8@giGt!rEA>So1fZ-;}1`=xo0*i&{;2oH@
zrE!)d(2-+A7<tvte!l}en2b}3e(-xGDs8O_Aj<|r*rx{OUQLqrCvhl{V~5k8@y`Lv
z;Ea91XEH}|hJg&DMl7e=<*0JOK8Z1l6>^Ioh_c@n2L1{3Mv0ZnRU8?MAPN`@+fajS
zw>eEqtY!d`1s`w|=Rsj7c!KVOIE!X}6cz%uCq-}|wEvs^DJY~IPxlzjaaTYvRtD=U
z_zCTUIDdAn=rAAgpxl2VDv^hU^1fM?F%uK%ZL56gREU^|RqSnxY$75MhuO*Hd;Cyz
zzrp=)pH(h%=U5_Y#$*28DcAY*p~#F^j`QAQzHTN-);src^F)HqJGb%o#`=jw>F-Yu
z9`kK8NwnU%iq{Xxka?xb@7%==7ddTu-5<pVHvLkbKF{M)D{E+TJT2D+14wFY$M1xB
z>qx3?abDviKd{lcQM+{fljZ9F1vKpc85(8|=Knr4NiueY^zflKpO7~Mik|uDao#H6
z5rYEb1{>uOs$Hbd#3GQO81bw>zC5_%ODeDeac0{XJwQogH7{Lx74N{lpOb4~jV^Y?
zuUR`I!&%b8G#`-8FOd2CF>B8uU0Isf1x5=~vX{<lJQ{bh6IM0)$~>r5dy40F5ZePK
zV6TrRAbvvEdGGh@DeEWXyvbWjk+{bhF!gx?$C2Ou789C1wYM2gwm@1{tS%u&jk@&e
z27Tt&+Lk7ue3!x3o%Dv~$y@f0avEzd0Ye{J8v;!WC+kv5sNSo_8yQ&-uZJLZq5%|7
zAu@iN4gY+0nAP~PqP;YRk+W^YxPC6`08zQK@2WXAG4&I;!VaaDA=i(|`>>xtI+Guw
zi$xMV@|XPD3ih3Giwq5l4K47vr_HJjCx_gsJ~=VzVsuuHTyr#FanPI&!}qlg)Ay+k
zgVTv-!ga2)A3^Q|_c`tp*@blx?BbpF(LH1?uh-$lMg!IH?IMs41Fr@1#&+>o#&-Q8
zxL-#)68nCB<5jE*$%XAWmK|~TsQVo{1^Eb_F=M|(gdKhlx*2Ax5f7efOrFYg@H!c;
zn2B6kI&70i9}le9+u`*PUvK$H7~4f};k9Jj(%2J7!n}OhoP$<u&d_|n$_*@N?R^0O
zOJj%q7psTkKd&BE4%Ywv>QVlWEf>D)e{H#V!~u{5YY++S)><i+O={xOY={!_;9=h%
zPgqEmsyy&AxjU1ccTi3*7cN|LDlIKYPoryDJZ@>-jY*d;*T`F2M>-xB)?D0LT5Fcv
zJbZO4-Bc~E`Y&swJB?4byQ5jn+`5!p^}eJS9||o2fxMUBTi-d!GD=XqYkKDg@-0ZP
z_g*~MIA<eH$NbMI#>tN(&nN9xdFk$l*<AWtsv0Z$(N4@=k_HP3n+3KN-{oOf1+<8X
zDlQz_JEt0~*SztaU)*rbYmH&FnzU<=Q#iR{zYS8n!V78L^*&p^wI4O}n{&^OoH#Vb
z^te^3mGxk1b5lQ^R@7PF><2Lq-yZeS^BuSC<JI*i3)PKZ)nB!9i4Q1Rs6&+XHcwa5
zj(PQb(VL47Z`g3f2Pr_%`z!#W?ELwe;XLtw_#<yANWp6V<U@?I|5=1@14H)jA5wJ3
zckvf)*M_%b{3D(g`lq~LRyCln4Uk^QQ(EU0y5F$Y$T>j*5C;PNpBb^fx$|5UefF@n
zXz*S%AU)gwMiCZLsM9%ieApher7<^jn#7-T*6fa_J5RHRyY>mcj_-45HK-zo(fGtd
zsx;g@qURJ;(7VYItHEtR7#P|J!*ZvbIL9;8kfKbGJ&<G6!IMzK8=F}5_wHb+9k!7s
z@f|iXoU;$q;E!%wL$p#}QdrCJ)%d1Em+XL=P7K8Nlskq^)r^;v{mlcA{icx!ITzxk
zJO@9uQ!FJq5(Y2rS9EtlYy$t%MJaM~@qcN`v#H0qK^>C+U}pkmYFq(VHZS^ktPnou
z8DEQ4Zlz`d^*CXD&xl`(7k1N@`{;7OdZM9yGemBwkl1=oC;rkw{8(QWY|N9>`P@81
z&`@b2sl&Ti*?n9t=ERHeX$Bu04{b;xe1SP!1r7dZ4mjxlD-QlgmJ>eyf1~vH{{{ym
zNngdhur6GCyWJ}Ta}K_orm#kO^r94g)_&C;&k+$~(``V@HYqMJrtD%16|k&0DLCiN
z=QW0_xbuYnQFshI{T>59pamM+n0e68lJPqIdz2`%v$^MS$|tcW#O!lMwXt#q*75Uq
zGRK;(bYV^q<Y3Nnh|<d*Wb1OKW4QiI4rrVFCXZ}_xC^SL{F2g(_eyf(W*rWThC{E;
zM4^&!Lv?4G_=VzQU{nStZRSp%Q<tL`Ul&O~7znsLVigGDPZ4;Yq2nK@Kbii~K>RQX
zUi<*$+SDOnieTbJ6CeeoK>p4a;Q5%4HYn)XRDYO?MixV6@Toz$MA0&cXA}GrJVgOR
zMbWyz2P`H{gwdzT|2)uV3$!yBu$N8)c%V<`iw+OiX8`Bh4zy|E)wr?Z)pUHeT+~e4
zH}PYJK~r$FJD%Nf80uLVz|VMhieA#p&h93XmvFAUk*C=zh?i@?>+0JIc-xXcZy$#j
z-$=DS`Z#Gv{azLmy6ps7Z!Sn(9nOvR^Gl*(G^BJ#FkZ*A2nzHt@CIyBJoG50d35o0
zy|Q1G#T36VUfe>u8J-jo8%9Fn0>}mt0No|Kf&)X7JtaVFEUr0PE0WJ|v0pze$?T`P
za5o!J`{-zb^fW<w9;si84|daJe`$d6wZQqD_qd%H-A&T~?qPIzi0Y<(PrRKOMs2@}
zzMp@w&6zF#%Mt*h(DK|oCA$To@mR>hPOH<|g*@jMaT0j?UtqxS{|gLC{|^TL2ZR3>
z1|55W_x}Y3jQ<$}HYSe$epU3Jg{Vrf@rlrW?f))BVZUB3?ugvip8sI-{hDK!eVbqI
zI&#WeHdDmjhnG{??M`|b63<Rf(AU@Z)tq?JT-;bnS)SQEJik>@#j@#dRYa-RG<B;#
zR?)nkXT?<CO=j^h*O(jVv-;KX#`6!hYd7htZ4FN5KmHBa2DTl$vrL<ZZ7nBh-b~(=
z7MkRwxgDo%3d}Dmxq1gE(si9{&&sS<8eRMl2Abnh!N<-^FlmWibZ{;W^zYsun9~Q!
zw_F4=>(8=sxM99Zm0sbc_1@PXWj^|@?UH}zu)Vo&VwLH0tJ7-OxHcC&eRp!8vwltQ
zrrdvfFiW@B;j5^qWi2=;W%w@mrsE%ZkImO>gO*hNWJ#G1xw0uDNgX+Phfr#Q{)47T
zWq{l?)ao@Z`$yjrQrpP>H*iP*6DkVn=vR(f1db645z)>H6@+POyVbRef{b}9D}Sv$
zqFr%o>tl!+QFuU8|APhLCaw%H3x?Vacirt}FjzUMZ_t9uQ50WEk{)E>MK|;9>}7vB
zfIisA`yQpRv)Y9?5eR070)TraM$(tiK1G@aT4B;Z5ipHC;2y=+1knlVCx{9@hjf><
z!Wr>6Xgiq2LB1u-W<%)f!PhH3NGhFlfZP|?zu$XK3ADv}FD|m03_$9?#Zi<4krn3=
zvKQneg5jAX0c1}KXGdUH5b!M^pd3T<9L|27jXzWd<0J5m<%KwNcKgSEHt0dQ=TulU
z69egEyEu6wD#FV~`P`-uF%zT6%NAMxH(yNWSB73c6h;2Z(EI->Lu-CzsMY_I4E+y<
z`r<Jk@+(7qav6VRC`4YV^gFl7|0YAZJbj+Zopv_B=6IU_Lx!$29>3$_brAiB3>_Wm
zfr-`~-=({ID3kv$konEQ#J4lFgyiP_U&)xi-VOc#3uK`GCo&u5$`xILhV2RX5rXWG
zJ2g9A5VRY({0Ap@579j79e8vb8P)4uIU#Y`UF%L>XHr~Je%I~)!@C({9iX$K(vtbV
zy_=S!W!OvJE#s^bTwH9Ah2#+_6SCzOVxyg3??xZb^~dyQa$Kp#^h3%A``mOm;9V6?
zOzh1xwGHt@+pu-F<!Z)dgPY-Ql*28g^WXe4Fz9rjH*()`Ql-$A@>f~#m4F&PamBeK
zy4b5u?xsZV{?&*_ZK;ftc@VL|gBqI$?mOJ#3;bl>$N00$!}K*DloQg1)Cvww_%AXw
z8YXbnc@JM!ZuQpRF&$I~SA`AI`|9x}^;MT`8!62@EH|v;?0OAF<jVmz<v+*BRjYWU
z<&MhT!8P(?7x)PU1B{`7ZmTHzjLgdF?Sks;*b04n_z@ukl19JZ3iSUJ%?CjqXJ>_K
ze%rxyk^q~skyHHKzRGgKsMLoG>XNxZf02Q77hnCO3qK9;AI=Rn(lGDAbSvtwb5lo_
z5pML_%l>qL63xUXYmn`?FPFXF49XpVD-3Q%&>O@h0iuqG_6YIVI|WGZ8BgIj8#KvZ
zk#99?KPsr81Ef2zd6>_ZOAvOA4@L*r6#@PUb+3Y-$6Y-7>)Ze&!EOGBb2CdvqGJTA
z{(rT1)<JQ#+nUE+g1fuZXk$$R0TSHZ-GUR`T>=RbG`PDYxNC5CZ#1~OOy@i2+&NPx
zHTT@AnLqB-R6#+%bno6&^;6HYetWHVV<D_jT}^i&7h52ue;)nmpJ5pdmHz&1IucMx
z-uKOKbhAg~3=Tp_R3f{Lyd}C+%ELKPcAiAuEe;v`^cBP}c-c6Cnn^--x4bvME6I{c
zQto5NV<r(qF5)Tm8{J$i3IBqio6xnLf$&NXE{JAi_byEFK??Ne9$XM~qXbeg|8-%i
z4pNYO5GLBZFjdZEh`$p~S>7GM1t~Dycu0qgsWAIwO!i)lo+ib%ySS&V<F$f3xARX{
zyuuaOKrW{}GoM=9`t+7MnIG4=K~E+fdyPoX(0Ss8zJJgR&%bI$fb+kmnI*wjEe9G^
zVNkz1c|~cJ+0kJ0Oko8sVdYLwD`I-Kki*Ty>oD(bVry;EK70|6o^fat5+V(C%k%|}
zdbj(<a9>D78k{wVD(I#2uGNKsk1JN(cJ)XV)=Umf{PaY8@k9%KZU_~?Z61~seH)ho
zbfVEKOO3H}WwG&M-ei60fgtcR)3eaQ$9^|Di?Q8}Bc5pPh^@kn%dc@DT>@lWzraAg
zC!E5~DvEZLD+%co|Eqca2eLjO{sk%?pgqousL!|Q6l=^-gOTk%TS%{MuYQS6r0VYM
z$vv_;?99x7HOcsyg^gTcO33k1ci2zXCq48`ZTaOr8=FW5W>`HEu^4`7;@+lAN%GE1
z+lmC>#i37Ubn1Z$MQ$z9!}S3~j=97xzjVfnFT<o^+_+E{3B2|zx9Zo)FsNvB9ZZzF
zpYT*FkIbKBQV;cCgkgkMg`3CrhhK+sZ--uohFi^n&dRX%h(fGFsG{nC9-oFv{Gm+<
zHQm9r4y7nUU7&4BIL)cs9VAt(teah*z{LI)kBfP9yvT?uSgi%Vhc&czdg<M1RhA(O
zf;H55H#<}VFBGEuX!YwoaMj+px(pl6I<$Kx-DUDJH)bkb)|)D=DSyxG(Nu(KB#;fU
zB28Is0rVAd3hxI_ja>WGg6_5q>+*`i<)`{@?4je@*)giLnou0YOm<8=G#a=`!S;+h
z@%A0c+N-EUv@$skJaIAFb!^69-Eu&|a+UNJ@{S_4)Y%u!X{EA2t5+OA#3q{oeg7_P
z$7n%q=jhf{fdtd6)N-*q)EK5UXW2Wp6L|jC9ru9bG3kJ`?6yh6N~4T{RwncO0ou?X
zanOL21{BkYNRtwAZ6fhI;aoV#jVh%FEiOEV?QWmn5&o<?!rK_u3h*09CEQX?uQUpY
zJgMws#WQFoPG`$Zf>k&GBJdCSknJzYkWGJIPLj24w3j(cJ!cuTD3=}vv$w$@N@;t*
zUNl9ipT?S$^8g#-PT*R$<~3aI^hN{~eGOwbqDjxHzdC8hII`P~0QCCN``1-oc8EBN
zG1e3lpH5t_+xZt>+J>KR)UMm>7hYD)ULFJ=M1LN>WYxR`jXq=2KZqtQyil3Ggwa2=
z13bPKw2Zui;4|Lm#!aTCR+1Hu)X8|U6TDVjcbZVZYR32OXMep*Df^A9v);K_xiirU
z&3*DeX!>989!>e*SJV7nAWZv0qxAP%AoRlego}xqcgeS~IN!H3fn;G%w`xTk%7bua
z{cQ^;oZZ#Q2JYc^i3>xV*Ha<U?g&4`tR^s5ew|CSudqi1jKrSwt1OE|b37O~;$9e7
zow=v3Mr`aFKb!1t@<dmHpE1p?w$HqQEfQ6)YBbB=QWz+A@;uxeu}9uX!LBN2tYRuz
z-m4FnEZv;B&wq4WjH#^Fly1Zx8ZlSvseZAM6IwlgbQ&pFsj!`losF_qV@;H==7O8X
ze~p|)-SHKQ6$-KA0UZjH@)RE(>PmG;N7$d&9~~OQzW|}AY^Dn;3*jp?qW?Y?V1~Lo
zD|CaHTv$30jMADf>sVd(S;;X}5*?MQ49rGp4oJ^ml?)7nht~lp!upecgzhMZ20A#!
zz>5)tNrcsDTc<!_c090V_itr*jI%<c2aZ70jOseelO=fyQ8%&CH(vCjxu0cLL5;*d
zQ#@qXSE@FyBHPF9M=pTiSrpei>P{H<_i~Oh0E)((23l`%K&ChB*U5AJpOjkIA-P{%
zrfWW6;p|#}0&It>Q7dqUVGBgY_T<DE<H-iOcM6Q_$y~$W&PJL{Po$b0&J1VM!p_*@
z4H`Je-es5!+Iw&`{Nn9L?eSJeSw_1`MDRktkUf?Iu*lH^$`aAS2YS>fqtr=N$R-B~
z2X~3+8i+EKltF28CLg<Rq`F>*eyBvU7)5!MneCx01bd^isCxVHnit<L&aB(swUD}N
z_X8TVjt>;q^!Kw0w%bgNhkSG*=MoH*KsIW|_-<!-K;YycN%EZD_dq{VoBYnsL~n+5
zy8gDDp%d%bfOA+cG>fh9*dGZ~PSMVeHZi{h?SKfG_xxk-?_F<0OAh@iYhPLio^Q;q
z+be5cT1TF5*dNyB7GB<%zaS<;?k#lf$GPy5J^YMG^&rZ#@bYP(mS_8eWYe}w>Dd_M
zMngvnd!thc`naCud-)wCfatU9EV66F!yHjh<jNEFHecJw7idMFU$TGDG~d7aG?0(~
z-_Eu^YH~K45J>)@VTl=X{aLbRvl1JohD;n|>1zS2mz-@riRIxsZPn9FDL=QJThe6E
z$kZNWthJU^&&Nww!FI`hg|=yCf8g8w(0If9J0EAXii(o9Mp9k3Hn$4bBcNB4O04)=
z*@~84QmOUK(HMV5%YN#d%z@<WZeCaGVKXMmtkT@r@oo?A(wd}=h!sZs)b?KJz=AA~
za>mWBhQd=|{$s+ut+Y}Ag&pwd;aEOJNAv`CsDkYQmRJa01*@<icX_4F22rV~Zk!}p
z@yEjsD~%^V;>oMHIs3v|p|U1Dis4=B`z;2A-HlI|t8i0c*{<DfUP^PB$d|!a+wR`3
zuX0_gG>c}By!B3oo_5x!FQY2fQpW713dM*?X6G-<g~<3nk?^fVq3&pNOt(}J1RN~B
z?8x>oEv@|2^Q@?NVwqNIL)t+YpwIJX!V1TF2Yhw++R`6ue9?sBr*Mo7hAHT?{}*Mm
zZ~k9`zU(3;<3?kbs_J%c8tMlqIZDI##ifqpY6dFhT8=RBM$Wz~mcgLRuXwdiOrHBx
zJ?6~^*|}{I&G-vJSl8k!cJmHP-7qs4QmhKYzznQ@V<Ix=78rR#o%IA~mG$V&o-u)X
zggVl&sit+;2o<LmsET=;X&$UyYCBEwH$b|W0?eK^$<rQ(CnRFDIR$KC!Xyfl`^lO9
z<bBo`aBTYh8lgK?3#tXx5&W!Cha8W|W&EI>zZ8+jhn64YBz>oXs6^sA_yEZ#+dg+>
zS$&doY8k>@->+Qs@_kR@;`I9CO<N%;ok?YeXd_w!gyd#;CM60_X@p+Xr9|QuqxU9d
zX|>OUkLg!a<bzs%Xd(g!3XDlt40|FLPfTZ8eq!FUDPqo}f-2sd>8_FqLxMDeX5T`6
zS+5sz2v|V5QY@hU4luW4k42eTuf?GZ3&=!|1=Ov;0>Y8<mON4S4)tXLi5KE!PBi0X
zZlvL57PsA^JwcX7#k~8PLT=G~sX##n;bY?};bRGq<K2jP)3@EqK1M4bbUEQ?Cy6~u
zafIB)`e=r~sK&qG$?)EZG8M(#V#@d0^qyve-AOk(Cup66Jt|7JgL80@rFhv>Zkn?i
z(Sq9P_vmw+DHxvHoKKQwszs@z$eXtNqVRLW1aD!NwRfTZAlW~(5&-#t{~aZ3&Nmgm
z!zrK_W0#yh@|I@X!_W6>nnQC?p?ItS7w>i>bgJ#-IH7Sy`=O)8X6V_!Prg&Nr}!76
zTm8x68{87MO9Xn=qc(w|@YRt{q2R5nrC)~_rFL_O7mcZ)yMZYWcbZoI&PyZ9m8ui{
z=_?Bu5hEV~as#p+Mlx3X5?3LExd*3q0U+;F&Vk#`#nMFa=e|&0KA!{me!0z1xf^g`
z<18v~<$=U{({;N5j^p4*ey602QkxeP*96E!$(tHvCPQgBV>oP#x1d+VhH&4n=MlML
zH&bC#Fli0*RJ9lkw+y!gVCdm<dzkU&PJzi9hN|hx0o8f*Jn&x)rx~@Kr9b8HD(}a^
zzAI*dYY6SP^gHCv{N%&m&KIP))tg<#7aT>zNjzti0wop9lALAHQ>`i4Pg?1Z(}xe-
zUd9D(0?(XB%LCZ@H@>sG>wyzUIs93!iFxIE8pBCu6<^6m<o<%KuEY`s=!ym+bYI28
zVfJnJC%3s=0SW|%+96#E&-@7bNGDxd@4NzB2_0}o2%qw5NY<4k^-1i@`$GXOqhfvI
zUnQcIMAHEJF!6JtEkCjQiZb(~BjO!j#y>|8d}sGCGC!qbAryC3=L(h5>P_BPd7<yi
z_-WlYo&u5U!A_Ydf~dvQ_b{hf3UP(edz?YDCuGktEy%dRZ04&>*Mjl%+ELs!pv5<=
z?V8bWC#}M4iYTADz23;Ny@|*%K-H}*=^2rY))~>mWaOCUT;$m4RODC>)?Yf!UW*(X
zx9YXHwClCFHSe`3;Q)!Z<7H;ozRIk&`<q_Z3BcT!xM1#W$njcVWkzs-a6~ub1FF&`
z_VQo7Ovq3@n^53`MQ67BdX?p#6;MiOo`Y1*v;p2lKyoQ}2c7{^ql{8(f$Y7vT}q#1
zcAnTR*}K;(k((m+mir^xPnI9dTn}S^=7iZbp5^rJWn(-;ox<-j{z0>U=wtx${rlIV
zAvXKFlcDv3r6o-Ii0TviMrkB95t1@Sb9t)uk2&0Qc4b(s_rIwtKlk_M(h|h+i`$Xg
zk?whoHS#*{C5$mU{1xk~J^5}k+vT#yHX#HJjhS&>j0({g2Z+t`Kx`IsbKLf~&8AsT
zx$TlbZV5DUknf6f;H<o%-}D?5=sJBB|G?`!;yzwHTf4(41d(j|w-H&I?P8&J&wT>c
zZO9$YRU%OyT{eqJ?uE?ke$0ZZj+dN(S=B0>1di6AffY9&p-8bSwAKJ*wIyB)qKRe!
z_pr|`K0@DCjHXV)3>#H1q(nUgW9{>9=#$rC%SoD_CG`C0$8qYCHqU0>L{D~F=WEyB
zxVCbLenc$XF>rYr_4X7J#G9P2>NB@o`PE>v+;Cu^TQ%*O|JAh>MR)C)--24+V^v}K
z$L{`OyBOJasiD<pTQQVm>;S@mVAwC1!3c!V<s#Ps!<!8|uXkXP37OPpkqHF|J!1Ud
zcQRH>_wov3e8nTgD^xWsD@K+J;3b4OY`Su;<)?(EEhy!E6zGZ#z+&e8FoDY2$jJ&L
zzK0`Y8Ty7XbabDqCVdCi1`7Qiyd&$^2|z!~dT=$oyZ}G^C{X?}-Py))$PrkMs-Fa;
z>qlLWY?YM5abyRF^5O=ofGzJ|AKJhvpv}c*bviuJb_O}#0SY)E^Wnv#_Z-hX#V?-%
z2)ddbuqO2l$q<&`a!ZO9A0oIXe3|RhIjA;5tE~MD_-=+}GChVLJ=d}IgNmaj7_e`_
zWCAu_7UeMo!<WnGC&POWp!AKWCq{25sRIh`r`sc{8ER4iE)cc$H3=NVc8+O@9%ZLP
zHldHNeGYcg@(Wybp3TJ0ykh9F_{PMuTgKhXKdNQ%wy}&mM|ak0uirS6&9fTJ?fD(-
zjK=!gQv0;kz5Cj}gSlsafVm$bKOxThTc&&UU~U)fGon_VGom)Fvy}a)v2lCIGAqQG
zb;8Gf%k>uGwjv?7vW<}Aar9V#qsDs1ZcQm!EwDV9di*Yx!(ZmZLT(j!?+mITRvRC<
zAO5ng+M<p|!5q_%MNTTmC6&BeqAkxT_{BkWGtQ$OHFXGhZxd)oEqLPyc5C%&FN^8M
z(z+aH<JbF6xH8vCp^3Q5{|CkXp?l=Nr(!v^544<<zp+d4kB*ubd^w-~W*~Bf`t@98
zzm`b?-nZ#&8vW@TSH15l;;fGj1|rW#lpSjM5r)Qx3VSJ~qKs^Q&TQ-E|H#GlzjLwe
z{nIPfB<trrNG?A6or^CXs$EUy+KUdx;5@soorK#2!<7Uz%T*7q+bV9h`;XVBAAjd!
zg@5E?G)OLX`$XdQ6ovX#yK=gwLL=Z{>91To_7Ks0t*iH-QIzYA$BG^&_YPWx`W4M*
z2g>t+aVVOsAN`=u9;9=mKfNiTY4QgJp=l6l#;`b<2<g;|Ar{(M1Ct9*K#At%&1R}X
z$d!a_kYkNo{zW;7^tH4eRR^!I<bz7)o$-G2T`Ng1H3vq>4=P<&|A_{SAvhqlZYNwT
zWlS%Z;Ep?dI%i!6({4crQGZ0HvmLsv07ju}=K*%%mOHd=^<0q^^{B*hRcN4tG^rd#
zxX-8ch>d$J9>R)DT6a!%(Z0+4(pbQZ)=O_D@TwhfD&|hWF1CnT*6p8W*|*1#xB?Z+
zb-w6sBf=NG;fl`{TE}V}eZLTRMx5Njw<YnFQ}T(`Og26GAuLq3CChaHwbq|QtCKLb
zHVz;>$lRw2DI5XUZ`HX&M<E*3<hOVN+zK2NiX=cB*VAb>Q<42ZVccu<UIh$M?`;`(
z$WFRz!EK+wK{R91U)dP_N%OE)f<xU~GFII?CZN{>96qKw6gf7Z1W|1hB=u&hd(+3<
zMz5CtEd<L$60aL1F7va1RD8~e?jg(HLT;m7QO}4T#k>*PUu9+&zRG0MfFx;%aJ|n`
zZizC{g11@4erIBRlrxHls#h<?^e-%5t<U|G70J#h(j(_0H^h6-fL%ek?{|<cbk&BC
zB^a<{taZ6g+}=17N%_f*%e~zN$v;-OL2|Jp@UL840LjG<kX#%guuJr>iUmOM^1ocC
z`EM%O?Ei1F|KDW)UozQIx?P$-N%jv-IsY~M)M@uuF2-9G&%cJ0Z%_sQYc96SL!gc$
zz5ktyt=u%!Ftc{g($3Ptsw>+>$`~q0GM;D~XZ9y3!Y=0T-(UGShgVeWXN#Qp_}y>#
z>^2V_p2>?ui#K00xayVftF;TQz@3-a8;pHxuL3b7KAX+?DOSZa`JT7aw~^h@#~$Ua
zD8%drQ7|xYPrUP3{H6~Jv00SA+wA>W!{383Rn=j$^?Yu%2@js5wda@Sm8%s-lo{?z
zNqY5_3Xy^E)gut_cP)FibyMmq^!9fx8#=IQ<qfH2CoX*tA0aXKcP$HR<O3qr7lG8W
zZ1^P?ON#ur=O(o`*7p)~C&PvHK^lNh=r*=tqZF8Fv$1K<y;9EQ1lFrOQ9o~1{G91G
z5_XS5n$dJuf;GzGjj=``J}V2YLiL&Yvm+&Pz&ylfT|1}@VdY6M=3o8n_*q;t+QHbt
zhyo{zAsM1(U4iAmQ`uQ0kdI*UUd|GQwbN4Z`$htS1W6tb9^&W8X(WeF{u9snW5yqU
zH9xfy_2y9i<+Gy@pN%O+15)XB!->$ubaM%AyQ8NAK6VJ~?sR<J?BM{`Lx23upSB?U
z$*~2lVgbURkpJRO?|}?|@hAG<{0R=ipT@X^2{nH6r>np5Cj;OJ#Al7&3C4{7=Cf81
zpG|`JEUU+dz#q^M{<PQ^ojpPyt+X~u$PqALBd$Lm4cK2w2RO3cV4|!4WwYEt9CfB>
zBX!1Tv9=?Bm9o7VRp=&rteSq2uBfw_fd*B+i{kJH3fYB*Pej`_&uFa#WugqAgR$^2
zL1#;LO^X<N-KtSYVzv-#ESnmS?F9*4f+5LwuV)um^Dj&Nk$xNP^9z!84|*(aAwMC`
zySEZCHh$7$vF8Fwyl!A_Z)Y&~3_)i08e~}uQuNm2W_EJ)TAa$95qV0Vr8q$fUVTUn
zOpqC2fA{xX+zZLYk&s-R|5q+{{GE&Ae&^!p7IhQ~>zG31zvtq8$$!elix<Gqf8^p5
z&lg`UvQXHXy=<LhrwL=1wYPpylT*8Pf70w9I*9&{+bq@ZzUs7;|7|Vn@lUmEI;6|M
z7dY2+&GnC3Hm_nx?#Hj@Uw!fXJ{O*LD<M|`#SR>LFK;>p_wNUuQX9?pqh~#5NzDr{
zCO*{;3`2-<)2|toc0(r7d{COQ>UtrG!pis>lk>r@kKs!Fw!>?aLPK`&^dq&~+SSot
ziiOap4|B3yId0CU7*rk)A5)kAsgzCa<Dc}x89wQs{iOGjQz448hrD!gcDqk0`OEN!
za{1Cj(`ph$bCAjfN0YAwsFmy+ij}RV3HSG7mt|hN<tq+$P3u<Wp(cI!3j3EdZmdG#
z6ORAOrEL3U3I{ucKDp)ogFeCk6MZ81KS`es+HVQupX||EUaf{ieZGb{;9RFq1>C<k
z<L}6gZsA+ijzu?d{}vLR%@zfL#C3)(2=tZc9$)AnC`yJ1hhg(Mn+k`+Qt!NrbzMfC
zGjylS6fL$U5-?NvEEg^EWhOeC8xo3VM<Afscl6#0bGuA5G$Z|rAz|nq=Q5-AIpMZU
zlnQhpn=JsTXWJY{|DweZ0wpglivM6c5e4DExp<i-hG!{2NC{g5=DyCy%dD<|pkW9i
zj-dG~4gZCjATVlg1p=i;BF9t>{sxgPs6Zwwe@n%kkot9xsmEdvay&085RQR2y>A}_
ztdT&}ID|aqLa1>+h_%PBWZ)$-KlL#aS5@i}u@L;j_*T(TIOcTSGESgNSYyUXY$75!
zmpEzoL2o?0f5HIA1!tl$7N`AS1wx-lWB#H~LlRLFK4g#3Z^d^p|0LNzbT9$Ax&Lis
z{I}_|{}_dcItm7=A`Og&O?UfR>XcujL@a$$#l~G9_iOj%4(=#H{QM+*@o4^dxa@Ox
zCtNGnepP#0irfB}EsFjlb9;^NVehzN?_g@wt^Z_|zl}%(GZwr%J~#HvczjU^F1Crq
zq`ECGT~qi(kqg3fdM-aIcy&l~Z19-5a9p**B(RjRcw@3OLK|qXxaoy-@1h5%xEvT5
z=lifRe!Y^bXpo-(&DczT;>Ni7YGh2g#~b%%@#BcYEKQvv0TWwl9Vo-%(%?i)^z-AR
zBk+Dj<z`W;G%@)FB`R$5^bIjZ&RuNwnfOv`1Y`ou=-1u%wRW$cueh`;B{mIUO^RL~
z%y)Yqr-(TK928z0JR^7qB@cZR*J+EGX~!c{EKGC8YH9~R-A4Rus&^8O4qVGrHxv$d
zbP!*|h9h%gzlJgF#G>=Z^{1lD4M*!ZkEB6VgsMtG_U~ekbYsHG?#K@BZi0t$D58Uv
zfx`DU!GMv5!t*zS$3Rye$xDHU%1=#LT!#M1c+U)zw+MiWArv&O$bb7bIw4tk8vAu1
zI{wdiuIOB~NU$=qJOL9FlFa+$(d=peZq6|AVEC{Zl_1u8oA}}Z{E`oXybEISBL(z2
zr&?aU0c79OQx9M7#}yA?FpbK~eJonuR*5GYbl9HWB9(Ubv2t=4w@eHnDpz?~LEHG8
zGLOp@^K1OBOjcktWmbK&1?;o!QK4y-8go(GI}Bv_mg-s2u!1}|or>x17&8SJlOiAS
zApSI$P7sYH|5Y0g24Fs?=_0t}Tn*oM_%Oc6337{bm{nJ1pDB0hB8z1L?E)Mp+am8|
zJ|cYYKg5pcey155T$fNznimYPQaR*x3ZDEahwf1oNTkdsDRyT2J-O^1NDnu-mTx~|
zgBZ^kZH_P(znhmH7xK;7YQ#W1ZFc*SaN4-Tz|>qx)ul5!x4%mXL(ek>d-O_XkfX8>
zuUBLI%fK&vZO@QA`dotgwwmqo%JinI#Hso=r=3|1_U5a^QhI}Jr*k!O1<&)XoOnAZ
z0rE<w)#o8`z06RmmW|7RtU{t#r=`0ge$%y+FA}2nx6uYd!@jB$q8~^m@c*R9KlDfO
z1OELsTp!3##{UGu0eyVI$;^>Gg$dntYAWT|sew}x)=<y_zgr5?-mCxS!6pzr?ko(C
z#O{8}oL))03i4voKb-#-I(WY@=5w;Lyrt(VxcJfYdtKd?hpYFiF>#HfLt4Aq4{aaP
zevN*q|8V74a+Gq#%k<+!&MV)L!tue{??Hu^G%;baGUnNR|7cwy_ngj2KkZkyV|y-g
zJ$;)ljw(2btgpa^BDl#7j(~Ad@E~UI!zF$#s2a!u;`tylOx;UD>m?o4R9tWuwK_AA
z61=st3tZT&6LDGfWW(pMb4(9aQ1;$s%LrY>kb7XaYggSPY*tzG(=5W8T(W)p0kfj&
zV#m?%yYZRcCc*xJg?&hYeJ^uqMV(1|<zbcmT6z6fCz@Ha#xq-;m-;$zbbYrh`fGIK
zo483M`wwDaBa^M)r;%t4X(M4au%SP_7GuI_e!bz)1b)2%!?d42>@W|{1&4xRns-n}
z895qRBKlx7%Wf82o)3<G1>AycA#|QM>TI~EVZmZ}Ov4nQGD_#$_lOHDlvf2;O-5&H
zNu06{B3CjVPm=O=Fo%-kaZ65>v^eqb94OLMXcdN<k7QU{PF0E9NI<URY}c?5KUh}~
zQS(DdArfE13+(3jDmswfi$}7QP$(aBiN_A%_0t&MzH(?Ef<#u7V+rV_XxOk^zsJ~c
z9EqR2aU#K)l6oTi_n!FlhA)nZ4hZzde%aksh2H&W1{{|r^C}Uoa!yyHd2^`dKRK>f
zPVeZ(KcMuT{;YCCzK*=p#hH!}UnRIVo{i6Xalc05HJQc9G9Pwh%oofnL>iY#$IkSO
zuvl|62Uuz1<62PM)o*aiJM<x6{5qWS;IMqB0@Aom8@=xwqZu6O<RD5ay{pV_hkM~Q
zpEPq)Ei`fv4?!S6M~)6F_>r_Q>m~w8nc42-Be$N+t?kFz(KUb4$3OQaC_34jx|%@7
zqZu7c%~j1b*#PW3oa|h@Oeh@UkojmAd&plP%l~+2WpDQuGDS_v_*RgMlZ%&AfJ=aj
zo12%Hn~jr)fs>N~@;o_v)Bo%#s!m1@4rZp1k!&MdXEPKIMO6t+c1c%TTVo?TyWi)c
zYH8(63HkclIk8fzn>jf{-kOq|9l!(N72x87ydf7m-#<?FuT!LywfbmA2{|inCQ2Iz
zBNt0&Gb756e}5q#J2yMze}asCo7tKE?l%Vf^Oasz4+k?!4h3Ut6_*bv9P*TWe7{Gx
zoSa=KdAWK1qc53<OW^Oe{+|zawd&}oki4()ovYmI7XcXU`$i?tBRSv98{(f5R(4~;
z5RX^9*2E_F@_QPPd9PPq@FCcOzKcDvq^6duy+s|iA}+B!uIzhRv}&Ss;(LUG#GAIs
zVib1u5@~5QVo8G(W-B!!_rCl9$HbaJwZ3i!cErcG_@`EKW^~?yYFdVerX^*eurOsL
zz_F0QMQI%(7F5<eSy<G^M8tcjGMK~eua2)VW9fnY1A_}iEQY`GgE^Yc@o#}TceHZb
z-gvmDuqi<RqjfX978q*15Tq_I9SP*HKx?dP7K+k6Y(}QiWL)83L2C_o#W736ozkrU
zk8m|<Np$Kx8oYEEE=FW2jfpPW01y_XVwS2Ink?ySD<~jdfaDaM;>_zZx<G_j8+ep1
zB54vfL==YE?%GoYuk~C<Y$&XdO8^WY%Z9cCwNKF>2MIp2kMbvHHF8KGvNW0$@wmHx
zY#tQ;ge6ml!DbJo-}i6u!Efx%u;}O<2Ofa8S91Ztr;i32U`uyB%%0gTJN>m0i$p{>
z0rK!oPd<-i{b}w~GHqZHhgA2-z}$XrFuW;#+%u1Vf8i-9d4TE5BZE_g|B5@8NF#;*
z^tyaiK>zUW)g^{^blb7_x07$fVHd=@m;?yM1hhrIPd~7m*F6J@KSB>A7s5&E^CF4s
z(^t92oc`)1g&w-dd)59}_O^i={k6iqI%PdDRi0gb-{T4^cxAP`$Lh&d-)E3n1mmlE
zsb%qHeYYC5BK6ed_t&{0li1MJs8m@=2MDdTQ5h(EU=O|AFf-4vn2cTIr{rBFBik~<
z(-gxrjQRT$#Aekd%-b#%z{EZS(+Zjz-En2*apBzVOn<(maq}}*Q&H><pD0yn>}M5z
zO4rG~*OBW_2nCw$(8#kPpFZz4+)TgRSrMlg2=Eoc+3-ObTdgN#Eh7Y%)>4LHV$HTr
z-FkC2$S2SXdh0Kt|Lg{JspyBH!y`AYOQxrYgCq*khi`}(FPLQCn@D3RFsZlX)PB{z
zh}MvCaTsfYt2v$b-hDFv(G#TcLtHKg<!2bJj$x93b;x&9JQK^c^?{ayR`^d@a9P_>
zI1g5*6Yz*8vYs@Z3<<{WeUgnv>uGWk>QpvDhG}II>TiO?6qo$f*21;?VDGdKp!vza
zDKH8p(B@9?(dUS`k9vu6n`iWfGYACD?qIb`MS<aK^i+fMlPnD=#`)1aBzQP2q)T}~
z6z_Q%7vAL1O9wvm$OhWpISN9hBSebkMGj|JS@xu^2utQ7F6*HN9k=Md>5~T=jSTyr
z92kp=N|cUwWhqQb!{Bm-(XpLauqZ8`wk(+NT4G!W$Drp*G4-%gA_=DwPFQ3cfKO!m
zF>rVUW%W5SdtBG8(E5lxC@GKBcl5J;vO>BhNa7g3@ZoFxta-C305}(HDuEU1to?kC
z?cp1q%#vkKgB=k=AQDO15$A3wh6t_U|M_bqb^zTrV|EAWU_vnF)(O^=|Jw%ihBu_C
zvqf2e#dlGdL$bK!8+npCf%5D117H*$+0pE^4J0nF(NaqhG~U-b;HEPa0U3$+nA76%
zB!LlGh&VxulGI*Mxa54FQ8@FEZS@FC0V>>HgGEIYqc54y^Ek<qr?>my7V~^cyS7@B
zvs=bu40e?o)G0z_A@XW>!9izXF8Ja>4BZCDp%k2u#EbC4f-dM$X!R^oEL^&*x8zbA
z>z0?ph<50p9Fx~m>|fsITEG>_WXU=u<iArSVMKW^9Cf_+g9eKyn(17L{vkFy9EnYh
zf(Z2|r@@j`AN)P(ns6M8(e`PCv1}X-7G*E#1*SzRJi0km{2U;&s3p+DYM?O7f_PXe
z)G2=G!ik_Ze!!Wy`l}N1AiZ=E`~gY2ElD-^uCRgN8?9+zFxpbqDP*J*>ILCPoRF`A
za(p*)^a2<CRa|hal{uZ76_f^$_XrOxLa({zxAG;h(J5pX&>b}X{v)e){_rNvj*M-+
zKlLISfS%d*J+Vlq+L?YkCv@dRZA$Z5!t<;7;Wa>d<2ZX;O(kDs#V86=xK^DcfTK*R
zlr3#PREjUeMuj`%y){kN3(KLC^LHr3pO*-^)5;oFJ~_;?OoF%hZXWN|_!eZ3U>w49
zKIP^G*u1u)%g7)N3aj3Cn6rt5Ik6(p$ccgDyHq6~SuftNq<u{+h$J$lM9xB``BAwJ
zo4z}1oG51}BeI0TMd&kgZlPtDw0ZVBqXebxI^3%*Y$iorar;;w(O;%#f)dBz@EAdB
z0;U=oWzL!i)_`!RlEYN2YIgCkegr%jyI(xlLF&#6v@r=`<qTsHjK#fnafg--=m-X;
zg5F_x#cR<~zL9~Q@8d5BuGLZIitAMT4LKxQQDY<S?a1}A+U6(O{CBnm$4tb8NcMxv
z&|Ty;Cli|KQ@7LIAY1baTWzTfVtJ-)|C(UfpD6+q)X_>xF!|dlFMpBV|DJu<mE@2y
zNIS1Ytqw8@%cpG?_+t2gh?bLkJfhfNP)w<a^ONTbX(WaQvvTHG+&CfQy)=*e=mDMG
zsKkCvv3K_q^BB+DRPmSXLJ!_tu>*Z$nhqA*(nr%bV?2C>O(_%aW?9^bxHG20iNb7u
zA-y37C=?4%e)yEbv*pn^>Kv_=X~l_eOrBEkqnVW%)lQU~i-P<v!Kp{N_A{F%4-Gzv
zbz{*q+#vc#$65v)Zn7Ti#MI=>YvQGktb89kBj$eHc?Z5hbXvf?3XG9t23GvoO0@3m
zaMoqp0E?JM(sh2{(KOVJq7P!d>wOE0k;WvdGtQ0vmE6Gq1&2?!F>3HMT1t=v=Ti<J
zM|W3l?xk_`P_0wq?iZlNKnleLeX7%(yRY1vCWpoQ2*5?&7X>D%!o8a0Zxrz#*1wA-
zzm;i4YK04cgL$J+%D&Blm%dTl;$=<MnulvbK>NjWf4ypGU0{c!yYE_c>eNS=3^hqe
z>N-b;>r=*ZLr=hNi#~N<^Uv8hJV(h7eDs~$`jRknKYi|mbiy(;mCt?c646^%Cr~-^
zMNT2p{RnM#XSlP#{;nNgYg3RgLTpMC=^a<P#yjTB+g`HyA%r#)rOBvEZFw_h+f*LC
z#0ZujKK?E!9+?8*+LZbNoyvTX#=IP?k`^7`jS0G8Haa-rT~~9hV=fca9rf}?`H0m-
zULGP0ZMSKyQ<PO@FnA#_!wQWMNy;|RL;pBOOgI;7Es0SUD^-`Uc{r0K1v56_p~3k3
ztYTjt-hdmEi;1^F<ca>kfb3M_1_@6*6FSkXGG-`g6-(}ojmsND;+<jA^Ys0;DNjd)
zsfpww3Kos3l6`!9$Vu#I9#b2Xih%+w1VfGaHv?=2*fRTWq#`q|ZpYis++lg&C{y=%
z)b<mfruQlIViKGtK7SNxPL4U=jb~~Hi;}K(jW#EF*<;rEE(p#Gt6=l>bsuP%r5MNi
z4xV=1?T7TC#+~1u>-Z9fC)J`pYQZkMBO2SdL$(yEuwS>igbnClUquZ`tK8{6+XHW}
zTs2hqKMl0BJ&_T1j%xWLp&*_%5Xqlez`wkTZ@z!Iy<EW(-jqF?il$(`*grho+rwk5
zohps4%aT364ANztY?|mBR3_3Yg8B9_mYEMXl~B!^%d4T4Y*Z?~d1&I$SNL(3b@-t2
z;%l58869RFs0FkLO^2y}#bo7Dm;C(<5ubalb{>@nmxb7+GDpYQ;E^7qSI3d|nJ%U`
z!kr@&Yz(s)DdR|2(R=l3cW%lPSXl^`j6PV9U5<=ss4UDK$Zw#CNU;>ho(<SqNxNxp
z^-%cSAcX~>f`#G#CW`!-)(iTslCa+8{lVV((!@sd+zY*4*XFERlmptn!X#aZzFu8t
zl?9tgOpl*pcU^0Xwp9O9>(tST&-huxSWvmHd(|UT5F6}Fp)cOSnKx3K7l=&%u|6aA
zp!CidB;l##=SvvVJF_Rlb}Z?}yTj-hEBb*>ITf(+bZbOkV9a;VyyGX^^RoXd4}Ma6
z!1iJN(!hR=`kjbx7J+>E_h%JOlqW)Tcs6b%jrcI8hR*VagD~TO&N+4z8?)w-`0_a~
zd#I3j%vL55H4esH@qobSV`5UMgkAXHBh2F@FcjCagW$Kv&eROd!Y%fzNyq&|zRi1l
zdF?HcYid~^;byXxfO-dnw-Nr}IN@;qf~%Q^XF)Gu=FHJM9Tsl#Zy%12xV?QAdZ0b!
z)ue^ItqNYyF7KbhU2KYL)t!We79Nirc9VkQM}k)E;4iDS)fY7%S1PEP%M2nSr}w6G
z)?}nkA=^|-ktRh3|KM65!nBa`qVN@1=&h5v7U1)&>omv@RIit&PsCcSlZ#V`bYSoz
zsZ27)VZP%l`oVJD!&0<o-!YYz>WiZGgm&K1*x)Y)^&ZLzS~ZBo@#5&YpiSp1{2&VW
z`K;`es5UqGQ&=qtfB+L5bqmYYQ;#2GLpKm#9~QP$NhD?S;GhcF+#`FIOn+c;e2aDi
z)kyRD=!scBLBJlW6oF|l3vNhv*uKD3&I%{XDxgMvR2a0lwzdGL7=vlL=gMvcg^Pr{
z?OGPc29rsh%b=Kn68Yum+)<uVH)yelhk2>)UR&#E>t?RGvi^PN^PUn`uGHp<b0<1|
z$~fcv+}_X?uT9tp2!7NC<y*J<X$wxf3Y5>Q7kVDgv`-gdXAK!aQ@XN!@y%wtt^!1{
z+2hkuf~*y%YvJjU*5}FJN*Smd?FUA$-A|Miu!78FV#<dt&hCu3#<ruW>)Xhxb!+cY
zNdwe;u}0W-w;S+Z4pd*N)Z6^p+ne3IdqQWQLO(9V15RFgGorutA*|n!E`DH{F-vn<
ziM;#@58{CFjhL;+P<AX2Jh!&lJ(>W};JF;WVb+ulz8(6G?`J~)l=!YSFWfy|WrIl{
z*_pb<PT1uOJb}M3B`zE8EHh$QrJH>07*xoV##01UqlCWyDnp7zv&_VqH$f<yLrOso
zISeTST)zkMy8sCTcVkU<#rw(5UtdtFo!tP9q$_$s*tw-FdmJaHH}6FEUk{kP71*>^
zC2If4E`X>?pBV6xESyDGh;;pEJH|?D0e5NP<mTcekCBkg$j9iYh_d0Iy{BpVHQi*r
z;I!;gJ5Qa{o~2oxNf^ev+3{`}CZ%S+kYS?)>fPvSX45lD#<>Hiu(AP3?rg_FeX4Tt
z<7(%RWs(;{f8P@~H=b`J6O}~}w9VSqLZLTmZ=PsG@v1Q1%?)aU&Rs%5>nVe$lw(3Y
z@2qt4t=2o7O=V&hp~;@Bd@Y})AJlTXiIOQ|#0{_bz`-#<D<4PoRh}RDXEijR(DH#b
zZ~9$&-aA1S_3_KYb2WM$i!n*3zDSmHj11a*!?O~!E7R_pmn2-nlC0#AeYhHkWF=#$
z%f}Hfoj}SJ>`J~0(ejC^rk3xI(wY3(C0|GLNzWiPF$v~_QQ%_+Cpu?=4PAX3<BHd5
zBoA3-=W0P!f_2pZBYO?^^Y&y(u&)ntGygCi^3x5Ss@$;9oe6U0m#XHuRy<^(<3urv
z(Bpn=5KV;JiOmOOlm1;79kG>n$-b-`+~=M*In~rg3`yYf7bUcB_u0WvUEKRlh1~{N
z)k_QwAgn8#x89P_dv$9p)X$73eAL7ba9Ej2`08tv6~WT|!)>f7)u-Ysc>RxLU!e8O
zkyET9gMI}|9uD5VYwTzjPTR=L5HUOueN`!yG(K{^=418D%^^;~^5JOq6}x!+0W>$`
z^XcIWG`~po{GZ6tzd{pDWixXWj`wz^X6}?aoRk1gpdJc`ij}9??_W_MHv`lGP;yfO
zAdAZO_AZdc--B+-zh7Z)4|(((Mg4yCmQqKUQ$n1JOB~1}&i__i04TsI@aC<++c(_&
z+~T}I-nZQ1Aj<#zDUkP(HM6sDv83b%{QK8n=thQEdW>-U&MqE(kLv-)Nkq{AH1p+?
zB!KCMp}<cG%7YY~`MAYQW&sHyAv#)`knNmpZ8fz=v}aZD@gpxt(-hWjrUfd)e^%0f
zN2nTUWBrld3U{SaZkh&1e=(XAzaL;EqV;jeoUFrVV0WYpVd0DV0zD~L@Q#CsT8}o(
z^o=*sQZg-8oC3-_Dkf?rWm)O)umo%lCYtTMK-D;AWjX72Y{fyrvq$xw?;d*x5w#w_
zpNq%6%s@2?)>sF(_-OLJX~R*6?#QgMMlgp_e`9FXSpu)wT7?2#OjfGCZZTSdK59G7
zjJA6|R6;ho8>EJ~d>^jHu-p^Uu_DW&^JDPi>jiX{0Xs8$tp!n4jM;?EWT!YK`LV=W
zk8NgyMqhP1zJ#SLCqIkb!VOT{!Nm-XANC$_>0dt~&MrnyF78f{j|3MF5Xi%iLPsZ|
IAc^un01Yy;$N&HU

literal 0
HcmV?d00001


From 5c775d89fcfbd3374c679019b03f0a1fd46eb46e Mon Sep 17 00:00:00 2001
From: Derek Smith <dkbsmith03@gmail.com>
Date: Sat, 1 Nov 2025 16:52:08 -0400
Subject: [PATCH 16/17] added env + main

---
 .../__pycache__/power_grid_env.cpython-38.pyc | Bin 0 -> 9909 bytes
 main/main.py                                  | 177 ++++++
 main/power_grid_env.py                        | 534 ++++++++++++++++++
 requirements.txt                              |   6 +
 4 files changed, 717 insertions(+)
 create mode 100644 main/__pycache__/power_grid_env.cpython-38.pyc
 create mode 100644 main/main.py
 create mode 100644 main/power_grid_env.py
 create mode 100644 requirements.txt

diff --git a/main/__pycache__/power_grid_env.cpython-38.pyc b/main/__pycache__/power_grid_env.cpython-38.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..7a3e059720f1efa36c075aa4c959fb183f8228a1
GIT binary patch
literal 9909
zcmb7KON<=HdG6QD^z_WmzJ|NY6(zR_hTM_J<&qLbY9WhDk)o_vT+<{a*)7wp-l^W*
zUd~I^JuA*+2n!Kq8$k%`5EzIpA-h2iCJ^M3lP`&{x#SS!)Q7-94$LAK2RSJS;Cz2o
zPtSv-kj$d1tLyQ1b^ZVU{g2vrCnhWfzt=y^20va?lz*d#;h%wrHC*8i3Z^&;Q<+v%
z9SwKArq*>wS7n>w7@{xZWN|lYxw`3?^}Lg>TaHyPIE8xADJEkloCz_u<doi0ScYZq
zD=h14JCmR4&J@#bDwSOO?^H#xuC;4XaO%>Q--zt%&5wO<zr_P)zuDLgc(YNb=FKLz
zZ}>q2_o^>i-tj%&2pU_Kb!$5aZGUKR7}VP}FY-g%vtPMzYNH+6EkSmR$bK9|+xFQN
z+oSh|tJW#|oW0>ik<SCnSmIv2Wpm74wa=~C*WN!}4;mE=x?pd4pi}cOX=K!ybd-%@
z=WXWKy#}+nAGVr}(6`%-U~I;@l&p0#!kg@{<yG;%GcT{4VgbG`44MtEW-rmoH@W{&
z+iz5{xRW-P>|<@%jAlD(wIejunjQ-))|*~+TYi$=+z5TX>xUqCZiHT~Sp`+u?L~Np
zb*V}>jKqFan~f-Fw43Q}Rkyvy7PjOc|Lpn?1F>yE13Qy!k6jDu_%Q1LBBQKV)Ho}?
zBCKR$<4RDCV&k14jN;6#cB|&c`lUuEHow!N*Lk%n%^CjXlZWtMQ1KN<g%Cpc9Gxxs
z2L3XB7FT9dJFE=VGJcNbzsUVW*;JXu3il@*lNH&-eZ|SMif{P_Mq7SC%qi|n_ywBJ
zCcn_x6q~-EaY{b5W<{CtC-GO<%wWXK7aE&o<@*N2^bnf^y=gYj7I2?ohuI?Tv+Nmm
z1otv~mOY32A@&Vs<37ipXGd|LXD_g0><{j%&H_8mPN3!RPZUFO79sR6#zi6IB;yeA
z?l-0Q3vnEl(_x8jhmdr8t5a`yVbHGIK?CBt387g&xPl!!D5H2J_waWYA8K2o`kQNO
zva+_m)^A(;i=^$Rf4lbRe;)qlKV93nelgaot#&0B=Nc{~&JSa=;Y#g@H5SEFeG%-q
z8_jlug|S6K=W-H{cuL9*B(vIaTfDguXPLhnRQ;+c7bHrmXL0GcLJLYCDW58P>b{~X
zk+!Gza{FpearGXRMo&YT>FFr5Jp*N~mqBUvvMBRR-2yt56{h2El$Cj43u;-aq2G$M
zodU}xW6X3+elTW7XVwFHN0Aj!POu`%66Q_P`(-c7;|&vn;#5!Bo2Her62{H2NtClu
zdFK#`5yodW)o5;Kp3OX<wJuN(o(~rkDybjW?4~v}W)T#h5fo=35M}Q_2nFrr_crh%
z@M=NVx1(+UDH5@q4o@qW_!iKtPlfIYJ6G-Xc6|e=+T64Se<42Oq?cZ~P#K$>;N#F_
z*<vr9T>&-@VdjeoIKBw<yK-Zc4vSXpTfj6sjQke87;xGKG>^P&T)<ZCs|o6LY$<Xy
z^IKi}gLAJgpF3&4va)>c_6UW6K(5;FKsbR@5KfwP?S1=$vnwkrCjsXxE2Fc<MR(P{
z4gvNyf?5!DXuH-=ok?&}$hY0XF4BjLyjzMp=5}tuI&H%Mkoen6C!kd)Y+&_?YMXf{
z2K(aVq`m1JWS|a?4#RRl3IfRE2RT%v+A43aJ;DQi$NH7>+C!Dkp(W{9PqbP2et(3r
zcJQa@KbS<3kon!;^8%hKMy%JGAIHXau(cg$!fmhR#~BuNTK;<F2pZb>y4ZBx00QW`
zaRz3lx-H(ur!b#SQ!zuuEER+~{1A$`AOyoDN@1LhV85C?w#ZTdBf#r;y3qvNy_%ac
z+Oamg(07Gdjtwx2#nS<78zkQaSHg(QY<yU(MJj_!#@mgWt!Ayc)p7C|68SF7Wt@XW
zqYBrs7)O7jc{kP?tvI&{p@?2O8y7uc<y>K(VjWI|Gd*MkllBY@wjWi@*a-dF=GT9*
z*p!EAoOj)7%?m@<eW?8Rw@<$dyBnTnKKJj0u#MZNuQaP|a#q6A)h6>#U%Gzg+=bJA
zi1HMWa>uVmr|VwOI4!n8ejrePxz*vv@hOYA!c`PX+0u$?cJQa$>PelR7j@IXT{l!s
z(EzZ<@6C*%;;vYt-aRyIYL@%Ftpab-AITs=^JQG&Cjy#x2nGR=;?kM=fS}lj2*|Wg
z)ho*FU)@t9WhcvM4E1YF2V5EimpcYPRNd5GQv{f5Oj}UyDE#Qy_zcEp=aqX}Pb;Le
zm^QD7nQ0rfB-B#M8Rf9j)6T<|;2ok>j*_#7dp|lV*UWN@idgskdwQho<(Nrd!t(q2
z543ygJzBAGN9Xa_8evkHHMT}$WPO+^o}a!pqlXzIYofO#E5ItuO=)seVLbAz0E1BM
zzWQ_{(ANN)oJ@~F&Y&xRo4ObA;O;U%0yZ8He>uqye};<Zs2~hDAWxrDZ$&*Ibx9!N
zw-?td7C(W8IMelcGnDdANE1&FUgFA^#2JCE{3LbfNO|Fp2zbC?@~X*=z}Qv`UP~pz
z&(SzFaI68PSaoA{hhIRCEMBGBGK%+bg~w1R#)4kbj_M_KK`rBNPPO$J+)X@B;cre|
z(7VS+<(r`ENpP`V(Mhj;2Cxu5C~%7VQJXia<dO7$IaE?rC}W#?DpT*|`4WT%LcOn&
zx3Z(}f$gF`s7Wgi{{*bn$hX)r%MeaC--9mh<iT<aZD~KL%Xw(5OFi&GcHn{ZXSx%p
zmmrw>Jp;Y)HU{(H4e1LCDC^QQLH{&^*8x=|<)n9$4>CjZ*x+qBv6k87U1>W;mPr-d
zLyY{cIilpw!#7#jSMGt{-eKIcz0BU?zVKIi*-eVt^4)((?cWUoP2o%oyQ@(XmRE2B
z))FxN1aIYr*!?kwWeAlKlU(B56INK5DtoE#l%Avzh5&zHQOVn(;BeKZ(1T(O-jRWx
z#~2nO!YL6wLW1#>u76$V%cDOk|IJ(f+P#0VuNox?0)HEY)Q`0vtzE4gk@`_Ubk~L{
zBWdC*7y?Ox1~lvZO?ohz4Sz^`_>0t7=-WL6aYA`mgMSkpiO!@^ots2;(zXiq%8+Lh
z@--qe$r1j2>}}Ni#7{2|8WI}#Hm9`PtVPg67cNi@FPs^6bI2C%`tjs&8?-gfM{v?y
z@@?X2cgv65VP3${QIps}axCIZqg`)xDmkIr;=}n>YBz-+23y~+d3WRD2qi-h8Pofh
zF?Sc2)O*uZEw!jx`XV&GtXn7-)H&41(q(m1HMQ={h~7)nt;IQM=|lb&h*b&S`ag27
z$j}KR_gAPu7ogG584WdI;_k_3L=<R+d4g#o-OZAzfu?53a0pF<vD=riiLz&s7Y|Lz
zCwD8k7oa&9vsc_#pvi{NrU{`<M6<*UK@;)B-XvOcz3ko;bPQ1jq8<a$PHr;;P0M({
z1_-6X`sQ7~3g6j=h}pz0=~Kf^r8r?o|DK?3j|T_IR(*|{Rip+<NfCL0WDZo#{@~P^
zllGZWcRF3@|3SIe{kxF_hv9_8Xjs4@?3}IIH@7|RQ$~dzMJmXv)sp;8`c8rX1jXcM
zMm{FJzwKjNxb3%s5YCs49MPbAbQW(ns!yNP?;V*V!RFG}gG)t|-ZNRV1W11Nn@4y+
zw(VzYoLsxgj6fA$L0t|$L$w#E5U%!jP>VD0GV87QS-0I{<n@M86Umt{3aW6H>qs+^
zj~SQU>b76KL;f=oITWxE+{~wwI?(RA%Q_9O9#r9!yT}GrMWW1=TN}?L6#5$lG~r~A
z`MXrSPX(zz{~n5pDV6^=>du7pV3Ujy*;ZmzZ1oAj@aCibM)D7c#t*0%#Q~UT;Ll&8
zBP8ENDV6mpy`<0RX7X3Wy`(|?r$^MEW{tSCWYbeS{>OL_+CFe+;gmwZq5aTBgCXlM
zue?e+4tu*tc2rtaq>qHYBPAsCovbnP3}}-sFRW^U5joI+qbg|R5*l#v$iaj~%`pqQ
zT<EU6E4T-#?{iNi?}Q_l0(n2hhpm^s+}1DtoPG~43I|RYC?Pq*JUx<~zrHA?3SY{9
zh+-YgIYAHWREXehj%wGbaH$Z%+YQtzx?rPI63D&l@jw_zXEt4tn>LQIi+_mu|BNdv
zqEPfXbw+(oFQe?fFv6cy0@Dv465GcZm%{WfK$ILW24`ze12kge5lOs53<`fvv4><B
zHlG|IV_%n}kS>Dzg)2vXH_N;#B8A2u(meiiVjxmX<Sgaj+I@~%UPJ+82=FdETg-$D
zr}b3fYC$;ACg{DUFwh1=bHWXJ{xa#C0HpCmfxk;Cxhyz!EAc8M+X8&mRq1wA3X)60
z&>I1?Q^eE!CWZ)$FIjhs9tG<J*NBq|*ZB9T%b{XalemYr-{J~sw4#;Otd>>hv^fzq
zTAHOEQ@hWPGU*B6F~XaE3Qw3&qCKuuhzV(5Ga}M@4tok07A95~(T^~shvD>6)b`O4
zF*5g*k!BBh+GONaVWjgHgt`hRy0D;-W4uQ$^F+_s*I-=LUIwk%o{UIK!h{#f$}?nI
zNy)7q$;2e$IX!W_El@yWPf&ylEmVD)^;O+sS#UVlJ=s@Tft_|AJ8hYd4L_!=EYkBr
zWu&^wSnFAA&p=hz`zSAbzEmCKybm{u^7%f$L=y<>`FE)x&6F&^Mm6C?X5qTxT!X((
zEvKju212MaC1esnb|xhj2Mx+E5Q4vf&NxRQS-|eb<-RZ4aVb$HmL57&ehs5}0Fxk-
zGAWt*hKDncs_%?U<iRdx?)u|$w}E&47p`ytg#zfFL6k<9l`X0zz;~IVwqv7`*SBh;
z25Ln3a-c?+5Wde*Az%ReDAGbiw69V^g5ol!&e9W<dCw5(i43(6<C7VgeU&oB7^5>o
z2oeb`(LoG#7^#Z%DT^oL4H}stpNs+T&qMS83>1AK*-;Q8x0l;jAS`D0yO;a+R5)aa
zT86w9N_R1MZmGD~XYHkw6(VOSSs~0pwHZcXWtqQ*FXi8&f?$Kw@!UYH1k@);AM4&m
z$OporhwZwQCy}eWgNc1_y@|Rroxa*_`N+&g9cOY7LI9p|bTb`#n|{=BcY`KQ(;!Kq
zGfC5i8n{m@8pEC{+9-Gq39b}yzoZ{Sh6?xJXSE^>c=x$c{*U`1Bb+DLNjd*A$PU8}
zc!l%G2RgW)CZ|4EZx_ifljnXIUO0pQji+2shmxz{o=oAu?h9@bv_GvUPZS=X14#mW
zq;Wf~05Tz@FoX69$cjPdoewCzlm!IAKVZ4fXf*^wc<xi}8O)qlKGQ@J3H5YEUpoTt
zN93?WXC~?Qpp7yGX&Zx#64j6Fo+^yh@wX7DKXw9$^F<`6{q*oK0WoZ|BFxnRxlvau
zTCCoQb2X3SJTO}4^JuJ0N#m8qvHUbm5`M--stHIF5-Dt#0Jj!8;;bjrM+Ahifqd4T
z*r@wnBi6%+_1#M0BNB4)1`0>ycnQAp45x~OM<FLHa3&i#2gafhoHAtH!*OAJ4}+FS
z7?)5{v^i}`wJZZdX{Of^B$bNlB3kGwtBa2b=y*IlBA}E~OMRA?@sI!w&MzgHFfD-*
zu#4$;3?ZBVpF0`YM+lBcs${w^NpB+9OhcA%yV<TWdjrv48raZzLh4B`Z_!67pY#v<
zK3QKA@DTjF4h9*LK_6j+L<aKW(Mst!XF7z7!u?@g-lTy<aeNOM+TNo*BwbX*E?HXl
z8)Hf=Vvqhpffgg1L;^)#qQGzBC}dTM;MU=;!@V>1>4|iB@Fen7X&*hA@(Hy(Mhj0I
zb&IJ9Zj&QuG3c4xaH7ieFUVOly5~v$pPUV{JsdJUp2d=>s{!5uKxqf#5((?zC?lR2
zU?~JRo{E|gtmptu5eA7C^01|E$3X_ppNOhXpw917A#{b1ns=xWd=pwD^hY>_1m-va
zA`Xr-iJ~))m?TlxP)o_5L|*f&?%XKD`Wc-#v)x8cYb0HwfmjN<e~kx`2-WznB1ITg
z^_R+*I0-;Hlng5!-~x#EaUA<g0N<B-(vwAx2&a;s9G;13{gLLE@D&A(JbDNk!C+>A
zMj@q9OnUN<^-!AB8d^t*Xt7B!d#Zc%hTp(B%y?qNUXs~guLj%t2|lZvy@}Xy)$UGR
z15eZcwdkNlUw&z`tK-nIJ99NXj1<7NYJbqxPTF1Lw&=Y@%pD5b#N0D9_uBj2qU<Hs
zqrH4N4b3~Y%*3Jf?5Y3vZ@p58Gc8VO_a_KZtWyR~uX}f&Bt*zcko*91kro+~B8uE;
z;~FGLgwKj*BGtaAzJO~HweIYQ5XsSxX*y}U&~(B^M)pVR&R(w37Ps-wHGbG^<KK9~
z<#qY`A7Rw^LXh%khJJ_ELBgnvr~E`WG%`<fEh-9B5XSHcDntr*619r9j(<-P?5;dR
z=bOlX_&DFRTo*ggMl#ini>?bVzLxalT$eSgXvSF~{)0uFBXWXxvQS){&55H5PAb9y
zwTesx=NinToHZva5E~oKW{q!C?}t=)RFHqnt5lF-;)JX+uksnyevBfXNDdXNey!%Z
zL;Lay4H4%ggu`Tm#p$4_8ak4YgTJpdRsLDXqoVfJ)N7`F{N&NIX3?BApEt|qvRN=E
ztO6%Gm1+JbxH<Xsze{i+-I0f>(tvyq-ASe!N4yAm<R=qv#jbGj5yW0Vro3pIf1Bv(
mZ#H(t3x>WxIEwTIvH6<Z``1ZSgrpf0Caf@$yXp}t_5TIdOsus4

literal 0
HcmV?d00001

diff --git a/main/main.py b/main/main.py
new file mode 100644
index 0000000..5ef24fc
--- /dev/null
+++ b/main/main.py
@@ -0,0 +1,177 @@
+"""
+Multi-Agent Reinforcement Learning for Power Grid Energy Flow Balancing
+
+This script demonstrates the usage of the PowerGridEnv environment
+and provides a simple test with random agents.
+"""
+
+import torch
+import numpy as np
+from power_grid_env import PowerGridEnv
+import matplotlib.pyplot as plt
+
+
+def test_random_agents(env, n_episodes=5, max_steps=100):
+    """Test the environment with random agents."""
+    print("Testing PowerGridEnv with random agents...")
+    
+    episode_rewards = []
+    episode_lengths = []
+    
+    for episode in range(n_episodes):
+        obs, info = env.reset()
+        episode_reward = 0
+        step_count = 0
+        
+        print(f"\nEpisode {episode + 1}:")
+        print(f"Initial state - Mean frequency: {info['mean_frequency']:.3f} Hz")
+        
+        for step in range(max_steps):
+            # Random actions for all agents
+            actions = np.random.uniform(-1, 1, size=(env.n_agents,))
+            
+            obs, reward, terminated, truncated, info = env.step(actions)
+            episode_reward += reward
+            step_count += 1
+            
+            # Print every 20 steps
+            if step % 20 == 0:
+                print(f"  Step {step}: Reward={reward:.2f}, Mean freq={info['mean_frequency']:.3f} Hz, "
+                      f"Safety violations={info['safety_violations']}")
+            
+            if terminated or truncated:
+                break
+        
+        episode_rewards.append(episode_reward)
+        episode_lengths.append(step_count)
+        
+        print(f"Episode {episode + 1} finished: Total reward={episode_reward:.2f}, Steps={step_count}")
+        if terminated:
+            print("  Terminated due to safety violations")
+        elif truncated:
+            print("  Truncated due to max steps")
+    
+    print(f"\nTest Results:")
+    print(f"Average episode reward: {np.mean(episode_rewards):.2f} ± {np.std(episode_rewards):.2f}")
+    print(f"Average episode length: {np.mean(episode_lengths):.1f} ± {np.std(episode_lengths):.1f}")
+    
+    return episode_rewards, episode_lengths
+
+
+def demonstrate_environment_features(env):
+    """Demonstrate key features of the updated environment."""
+    print("\n" + "="*60)
+    print("POWER GRID ENVIRONMENT DEMONSTRATION")
+    print("="*60)
+    
+    # Reset environment
+    obs, info = env.reset(seed=42)
+    
+    print(f"\nEnvironment Configuration:")
+    print(f"  Number of buses: {env.n_buses}")
+    print(f"  Number of agents: {env.n_agents}")
+    print(f"  Agent types: {env.agent_types}")
+    print(f"  Ramp rates (MW/min): {env.ramp_rates.cpu().numpy()}")
+    print(f"  Power limits (MW): min={env.power_min.cpu().numpy()}, max={env.power_max.cpu().numpy()}")
+    print(f"  Cost coefficients ($/MW): {env.cost_coefficients.cpu().numpy()}")
+    print(f"  Frequency bounds: {env.frequency_bounds} Hz")
+    print(f"  Load range: {env.load_range} MW")
+    print(f"  Communication delay: {env.delay_steps} steps")
+    print(f"  Observation dimension: {env.obs_dim} per agent (15 as specified in proposal)")
+    
+    print(f"\nInitial State:")
+    print(f"  Observation shape: {obs.shape}")
+    print(f"  Action space: {env.action_space}")
+    print(f"  Mean frequency: {info['mean_frequency']:.3f} Hz")
+    print(f"  System freq deviation: {info['system_freq_deviation']:.4f} Hz")
+    print(f"  Total load: {info['total_load']:.1f} MW")
+    print(f"  Total generation: {info['total_generation']:.1f} MW")
+    print(f"  Current time: {info['current_hour']:.1f}h, Day {info['current_day']:.0f}")
+    
+    # Test full state vector
+    full_state = env.get_full_state()
+    print(f"  Full state shape: {full_state.shape} (for centralized critic)")
+    
+    # Test different action scenarios
+    print(f"\nTesting Action Scenarios:")
+    
+    # Scenario 1: All agents increase output (respecting capacity)
+    print("\n1. All agents increase output:")
+    actions = np.ones(env.n_agents) * 0.5
+    obs, reward, terminated, truncated, info = env.step(actions)
+    print(f"   Reward: {reward:.2f}")
+    print(f"   Mean frequency: {info['mean_frequency']:.3f} Hz")
+    print(f"   System freq deviation: {info['system_freq_deviation']:.4f} Hz")
+    print(f"   Total generation: {info['total_generation']:.1f} MW")
+    print(f"   Capacity utilization: {[f'{u:.2f}' for u in info['agent_capacity_utilization'][:5]]}")
+    
+    # Scenario 2: Test capacity constraints
+    print("\n2. Test capacity constraints (large actions):")
+    actions = np.ones(env.n_agents) * 1.0  # Maximum actions
+    obs, reward, terminated, truncated, info = env.step(actions)
+    print(f"   Reward: {reward:.2f}")
+    print(f"   Mean frequency: {info['mean_frequency']:.3f} Hz")
+    print(f"   Safety violations: {info['safety_violations']}")
+    print(f"   Capacity utilization: {[f'{u:.2f}' for u in info['agent_capacity_utilization'][:5]]}")
+    
+    # Scenario 3: Mixed actions by agent type
+    print("\n3. Mixed actions by agent type:")
+    actions = np.zeros(env.n_agents)
+    actions[:5] = 0.3    # Batteries moderate increase
+    actions[5:13] = -0.2 # Gas plants decrease
+    actions[13:] = 0.1   # DR slight increase (load reduction)
+    obs, reward, terminated, truncated, info = env.step(actions)
+    print(f"   Reward: {reward:.2f}")
+    print(f"   Mean frequency: {info['mean_frequency']:.3f} Hz")
+    print(f"   System freq deviation: {info['system_freq_deviation']:.4f} Hz")
+    print(f"   Total generation: {info['total_generation']:.1f} MW")
+    
+    print(f"\nAdvanced Features:")
+    print(f"  Communication delay active: observations are {env.delay_steps} steps delayed")
+    print(f"  Renewable forecasts: {env.renewable_forecasts.shape}")
+    print(f"  Safety violations: {info['safety_violations']}")
+    print(f"  Contingency active: {info['contingency_active']}")
+
+
+def main():
+    """Main function to run the power grid environment demonstration."""
+    # Set device
+    device = 'cuda' if torch.cuda.is_available() else 'cpu'
+    print(f"Using device: {device}")
+    
+    # Create environment
+    env = PowerGridEnv(device=device)
+    
+    # Demonstrate environment features
+    demonstrate_environment_features(env)
+    
+    # Test with random agents
+    episode_rewards, episode_lengths = test_random_agents(env, n_episodes=3, max_steps=50)
+    
+    # Plot results
+    plt.figure(figsize=(12, 4))
+    
+    plt.subplot(1, 2, 1)
+    plt.plot(episode_rewards, 'o-')
+    plt.title('Episode Rewards')
+    plt.xlabel('Episode')
+    plt.ylabel('Total Reward')
+    plt.grid(True)
+    
+    plt.subplot(1, 2, 2)
+    plt.plot(episode_lengths, 'o-')
+    plt.title('Episode Lengths')
+    plt.xlabel('Episode')
+    plt.ylabel('Steps')
+    plt.grid(True)
+    
+    plt.tight_layout()
+    plt.savefig('/Users/dereksmith/Documents/code/AP158/es158-project/test_results.png', dpi=150)
+    plt.show()
+    
+    print(f"\nEnvironment test completed successfully!")
+    print(f"Results saved to: test_results.png")
+
+
+if __name__ == "__main__":
+    main()
diff --git a/main/power_grid_env.py b/main/power_grid_env.py
new file mode 100644
index 0000000..2aeb371
--- /dev/null
+++ b/main/power_grid_env.py
@@ -0,0 +1,534 @@
+"""
+Multi-Agent Power Grid Environment for Reinforcement Learning
+
+This environment simulates a 68-bus power grid with 20 agents:
+- 5 batteries (ramp rate: 50 MW/min)
+- 8 gas plants (ramp rate: 10 MW/min) 
+- 7 demand response units (ramp rate: 5 MW/min)
+
+State space: 140-dimensional (bus frequencies, generator outputs, loads)
+Each agent observes: 15-dimensional local observation
+Actions: 20 continuous power changes ΔP^i within ramp rate limits
+"""
+
+import gymnasium as gym
+from gymnasium import spaces
+import torch
+import numpy as np
+from typing import Dict, List, Tuple, Any, Optional
+import math
+import random
+
+
+class PowerGridEnv(gym.Env):
+    """Multi-agent power grid environment using gymnasium interface."""
+    
+    def __init__(self, 
+                 n_buses: int = 68,
+                 n_agents: int = 20,
+                 dt: float = 1.0,  # time step in minutes
+                 frequency_bounds: Tuple[float, float] = (59.5, 60.5),
+                 load_range: Tuple[float, float] = (2000.0, 5000.0),
+                 contingency_prob: float = 0.001,
+                 device: str = 'cpu'):
+        """
+        Initialize the power grid environment.
+        
+        Args:
+            n_buses: Number of buses in the power grid (68)
+            n_agents: Number of agents (20: 5 batteries + 8 gas + 7 DR)
+            dt: Time step in minutes
+            frequency_bounds: Frequency bounds in Hz [59.5, 60.5]
+            load_range: Load range in MW [2000, 5000]
+            contingency_prob: Probability of N-1 contingency per step
+            device: PyTorch device ('cpu' or 'cuda')
+        """
+        super().__init__()
+        
+        self.device = torch.device(device)
+        self.n_buses = n_buses
+        self.n_agents = n_agents
+        self.dt = dt
+        self.frequency_bounds = frequency_bounds
+        self.load_range = load_range
+        self.contingency_prob = contingency_prob
+        
+        # Agent configuration: [batteries, gas plants, demand response]
+        self.agent_types = ['battery'] * 5 + ['gas'] * 8 + ['dr'] * 7
+        self.ramp_rates = torch.tensor([50.0] * 5 + [10.0] * 8 + [5.0] * 7, device=self.device)  # MW/min
+        
+        # Agent capacity constraints [P_min, P_max] in MW
+        self.power_min = torch.tensor([0.0] * 5 + [50.0] * 8 + [-200.0] * 7, device=self.device)
+        self.power_max = torch.tensor([100.0] * 5 + [500.0] * 8 + [0.0] * 7, device=self.device)
+        
+        # Agent-specific cost coefficients ($/MW)
+        self.cost_coefficients = torch.tensor([5.0] * 5 + [50.0] * 8 + [20.0] * 7, device=self.device)
+        
+        # Wear-and-tear coefficients for different agent types
+        self.wear_coefficients = torch.tensor([0.1] * 5 + [0.05] * 8 + [0.2] * 7, device=self.device)
+        
+        # Power system parameters
+        self.nominal_frequency = 60.0  # Hz
+        self.base_power = 100.0  # MVA base
+        self.inertia_constants = torch.rand(n_buses, device=self.device) * 5.0 + 2.0  # H in seconds
+        
+        # Grid topology - simplified admittance matrix (68x68)
+        self._initialize_grid_topology()
+        
+        # State space: [frequencies (68), generator outputs (20), renewables (14), loads (30), time features (8)] = 140
+        self.state_dim = 140
+        self.obs_dim = 15  # Local observation dimension per agent as specified in proposal
+        
+        # Action space: continuous power changes for each agent
+        self.action_space = spaces.Box(
+            low=-1.0, high=1.0, shape=(self.n_agents,), dtype=np.float32
+        )
+        
+        # Observation space: each agent gets local 15-dim observation
+        self.observation_space = spaces.Box(
+            low=-np.inf, high=np.inf, shape=(self.n_agents, self.obs_dim), dtype=np.float32
+        )
+        
+        # Communication delay buffer (2-second SCADA delay)
+        self.delay_steps = 2  # 2 time steps delay
+        self.observation_buffer = []
+        
+        # Renewable forecast parameters
+        self.forecast_horizon = 5  # 5 time steps ahead
+        self.renewable_forecasts = torch.zeros(14, self.forecast_horizon, device=self.device)
+        
+        # Initialize state variables
+        self.reset()
+    
+    def _initialize_grid_topology(self):
+        """Initialize the power grid topology and admittance matrix."""
+        # Simplified 68-bus system admittance matrix
+        # In practice, this would be based on actual grid topology
+        self.admittance_matrix = torch.zeros(self.n_buses, self.n_buses, device=self.device)
+        
+        # Create a connected graph structure
+        for i in range(self.n_buses):
+            # Self admittance (diagonal elements)
+            self.admittance_matrix[i, i] = torch.rand(1, device=self.device) * 10.0 + 5.0
+            
+            # Connect to nearby buses (simplified ring + radial structure)
+            if i < self.n_buses - 1:
+                admittance = torch.rand(1, device=self.device) * 2.0 + 1.0
+                self.admittance_matrix[i, i+1] = -admittance
+                self.admittance_matrix[i+1, i] = -admittance
+                self.admittance_matrix[i, i] += admittance
+                self.admittance_matrix[i+1, i+1] += admittance
+        
+        # Add some additional connections for robustness
+        for _ in range(self.n_buses // 4):
+            i, j = torch.randint(0, self.n_buses, (2,))
+            if i != j:
+                admittance = torch.rand(1, device=self.device) * 1.0 + 0.5
+                self.admittance_matrix[i, j] = -admittance
+                self.admittance_matrix[j, i] = -admittance
+                self.admittance_matrix[i, i] += admittance
+                self.admittance_matrix[j, j] += admittance
+        
+        # Agent-to-bus mapping (which buses have controllable agents)
+        self.agent_bus_mapping = torch.randint(0, self.n_buses, (self.n_agents,), device=self.device)
+    
+    def reset(self, seed: Optional[int] = None, options: Optional[Dict] = None) -> Tuple[np.ndarray, Dict]:
+        """
+        Reset the environment to initial state.
+        
+        Returns:
+            observations: Array of shape (n_agents, obs_dim)
+            info: Dictionary with additional information
+        """
+        if seed is not None:
+            torch.manual_seed(seed)
+            np.random.seed(seed)
+            random.seed(seed)
+        
+        # Initialize bus frequencies around nominal (60 Hz)
+        self.frequencies = torch.ones(self.n_buses, device=self.device) * self.nominal_frequency
+        self.frequencies += torch.randn(self.n_buses, device=self.device) * 0.01  # Small initial deviation
+        
+        # Initialize generator outputs (MW) within capacity bounds
+        self.generator_outputs = torch.zeros(self.n_agents, device=self.device)
+        # Set initial outputs to minimum capacity for gas plants and mid-range for batteries
+        self.generator_outputs[:5] = 50.0  # Batteries at 50% capacity
+        self.generator_outputs[5:13] = 100.0  # Gas plants at minimum
+        self.generator_outputs[13:] = -50.0  # DR at 25% load reduction
+        
+        # Initialize loads (MW) - random within specified range
+        load_min, load_max = self.load_range
+        self.loads = torch.rand(self.n_buses, device=self.device) * (load_max - load_min) + load_min
+        
+        # Initialize renewable generation (14 buses with renewables)
+        self.renewable_buses = torch.randint(0, self.n_buses, (14,), device=self.device)
+        self.renewable_generation = torch.rand(14, device=self.device) * 500.0  # 0-500 MW
+        
+        # Initialize voltage angles (radians)
+        self.voltage_angles = torch.zeros(self.n_buses, device=self.device)
+        
+        # Reset contingency state
+        self.contingency_active = False
+        self.contingency_bus = None
+        
+        # Time step counter and time features
+        self.time_step = 0
+        self.current_hour = 12.0  # Start at noon
+        self.current_day = 1.0    # Monday
+        
+        # Initialize observation buffer for communication delay
+        initial_obs = self._get_observations_immediate()
+        self.observation_buffer = [initial_obs.clone() for _ in range(self.delay_steps + 1)]
+        
+        # Initialize renewable forecasts
+        self._update_renewable_forecasts()
+        
+        # Get initial observations (with delay)
+        observations = self._get_observations()
+        info = self._get_info()
+        
+        return observations.cpu().numpy(), info
+    
+    def step(self, actions: np.ndarray) -> Tuple[np.ndarray, float, bool, bool, Dict]:
+        """
+        Execute one time step of the environment.
+        
+        Args:
+            actions: Array of shape (n_agents,) with continuous actions in [-1, 1]
+            
+        Returns:
+            observations: Next state observations
+            reward: Shared reward for all agents
+            terminated: Whether episode is terminated
+            truncated: Whether episode is truncated
+            info: Additional information
+        """
+        actions = torch.tensor(actions, device=self.device, dtype=torch.float32)
+        
+        # Convert normalized actions to actual power changes (MW)
+        power_changes = actions * self.ramp_rates * self.dt  # Scale by ramp rate and time step
+        
+        # Apply action constraints - check capacity limits before updating
+        feasible_changes = torch.zeros_like(power_changes)
+        for i in range(self.n_agents):
+            current_power = self.generator_outputs[i]
+            desired_change = power_changes[i]
+            
+            # Calculate feasible change respecting capacity bounds
+            max_increase = self.power_max[i] - current_power
+            max_decrease = self.power_min[i] - current_power
+            
+            feasible_changes[i] = torch.clamp(desired_change, max_decrease, max_increase)
+        
+        # Apply feasible power changes to generator outputs
+        self.generator_outputs += feasible_changes
+        
+        # Ensure outputs stay within bounds (safety check)
+        self.generator_outputs = torch.clamp(self.generator_outputs, self.power_min, self.power_max)
+        
+        # Store action for wear calculation
+        self.last_actions = feasible_changes
+        
+        # Update time features
+        self._update_time_features()
+        
+        # Update stochastic loads and renewable generation
+        self._update_stochastic_components()
+        
+        # Update renewable forecasts
+        self._update_renewable_forecasts()
+        
+        # Check for N-1 contingencies
+        self._check_contingencies()
+        
+        # Solve power flow and update frequencies using swing equation
+        self._update_system_dynamics()
+        
+        # Calculate reward
+        reward = self._calculate_reward()
+        
+        # Check termination conditions
+        terminated, truncated = self._check_termination()
+        
+        # Update observation buffer and get delayed observations
+        current_obs = self._get_observations_immediate()
+        self.observation_buffer.append(current_obs)
+        if len(self.observation_buffer) > self.delay_steps + 1:
+            self.observation_buffer.pop(0)
+        
+        # Get delayed observations
+        observations = self._get_observations()
+        
+        # Update time step
+        self.time_step += 1
+        
+        info = self._get_info()
+        
+        return observations.cpu().numpy(), reward, terminated, truncated, info
+    
+    def _update_stochastic_components(self):
+        """Update stochastic load and renewable generation."""
+        # Add random variations to loads (±5% per step)
+        load_variation = torch.randn(self.n_buses, device=self.device) * 0.05
+        self.loads *= (1.0 + load_variation)
+        self.loads = torch.clamp(self.loads, self.load_range[0], self.load_range[1])
+        
+        # Update renewable generation with stochastic variations
+        renewable_variation = torch.randn(14, device=self.device) * 0.1
+        self.renewable_generation *= (1.0 + renewable_variation)
+        self.renewable_generation = torch.clamp(self.renewable_generation, 0.0, 1000.0)
+    
+    def _check_contingencies(self):
+        """Check for N-1 contingency events."""
+        if torch.rand(1).item() < self.contingency_prob:
+            if not self.contingency_active:
+                # Activate contingency - disconnect a random bus
+                self.contingency_active = True
+                self.contingency_bus = torch.randint(0, self.n_buses, (1,)).item()
+                # Reduce load at contingency bus
+                self.loads[self.contingency_bus] *= 0.1
+        else:
+            # Recover from contingency
+            if self.contingency_active:
+                self.contingency_active = False
+                if self.contingency_bus is not None:
+                    # Restore load
+                    load_min, load_max = self.load_range
+                    self.loads[self.contingency_bus] = torch.rand(1, device=self.device) * (load_max - load_min) + load_min
+                self.contingency_bus = None
+    
+    def _update_system_dynamics(self):
+        """Update system dynamics using swing equation."""
+        # Calculate power imbalance at each bus
+        power_injection = torch.zeros(self.n_buses, device=self.device)
+        
+        # Add generator outputs to their respective buses
+        for i, bus_idx in enumerate(self.agent_bus_mapping):
+            power_injection[bus_idx] += self.generator_outputs[i]
+        
+        # Add renewable generation
+        for i, bus_idx in enumerate(self.renewable_buses):
+            power_injection[bus_idx] += self.renewable_generation[i]
+        
+        # Subtract loads
+        power_injection -= self.loads
+        
+        # Simplified swing equation: df/dt = (P_mech - P_elec) / (2 * H * f_nom)
+        # where P_elec is calculated from power flow
+        
+        # Calculate electrical power flow (simplified)
+        frequency_deviations = self.frequencies - self.nominal_frequency
+        electrical_power = torch.matmul(self.admittance_matrix, frequency_deviations)
+        
+        # Power imbalance
+        power_imbalance = power_injection - electrical_power
+        
+        # Update frequencies using swing equation
+        frequency_derivative = power_imbalance / (2.0 * self.inertia_constants * self.nominal_frequency)
+        self.frequencies += frequency_derivative * self.dt * 60.0  # Convert minutes to seconds
+        
+        # Enforce frequency bounds
+        self.frequencies = torch.clamp(self.frequencies, self.frequency_bounds[0], self.frequency_bounds[1])
+    
+    def _calculate_reward(self):
+        """Calculate the shared reward based on equation 2 with exact coefficients."""
+        # 1. Frequency deviation penalty: -1000 * sum of squared deviations
+        frequency_deviations = self.frequencies - self.nominal_frequency
+        frequency_penalty = 1000.0 * torch.sum(frequency_deviations ** 2)
+        
+        # 2. Agent-specific costs: C_i per MW adjusted (based on last actions)
+        if hasattr(self, 'last_actions'):
+            agent_costs = torch.sum(self.cost_coefficients * torch.abs(self.last_actions))
+        else:
+            agent_costs = 0.0
+        
+        # 3. Wear-and-tear functions: 0.1 * W_i (based on action magnitude)
+        if hasattr(self, 'last_actions'):
+            wear_costs = 0.1 * torch.sum(self.wear_coefficients * (self.last_actions ** 2))
+        else:
+            wear_costs = 0.0
+        
+        # 4. Safety constraint violations: exactly 10,000 per violation
+        safety_violations = 0.0
+        freq_violations = torch.sum((self.frequencies < self.frequency_bounds[0]) | 
+                                  (self.frequencies > self.frequency_bounds[1]))
+        safety_violations = freq_violations * 10000.0
+        
+        # Total reward (negative because we minimize costs)
+        reward = -(frequency_penalty + agent_costs + wear_costs + safety_violations)
+        
+        return reward.item()
+    
+    def _get_observations(self):
+        """Get delayed observations for each agent (2-second SCADA delay)."""
+        if len(self.observation_buffer) >= self.delay_steps + 1:
+            return self.observation_buffer[-(self.delay_steps + 1)]  # Return delayed observation
+        else:
+            return self.observation_buffer[0]  # Return most recent if buffer not full
+    
+    def _get_observations_immediate(self):
+        """Get immediate (non-delayed) observations for each agent."""
+        observations = torch.zeros(self.n_agents, self.obs_dim, device=self.device)
+        
+        # Calculate system frequency deviation (key coordination signal)
+        system_freq_deviation = torch.mean(self.frequencies - self.nominal_frequency)
+        
+        for i in range(self.n_agents):
+            bus_idx = self.agent_bus_mapping[i]
+            obs_idx = 0
+            
+            # Local bus frequency (1)
+            observations[i, obs_idx] = self.frequencies[bus_idx]
+            obs_idx += 1
+            
+            # Local bus load (1)
+            observations[i, obs_idx] = self.loads[bus_idx]
+            obs_idx += 1
+            
+            # Own generator output (1)
+            observations[i, obs_idx] = self.generator_outputs[i]
+            obs_idx += 1
+            
+            # System frequency deviation Δf_sys = (1/68)Σ(f_k - 60) (1)
+            observations[i, obs_idx] = system_freq_deviation
+            obs_idx += 1
+            
+            # Nearby bus frequencies (5 nearest buses)
+            distances = torch.abs(torch.arange(self.n_buses, device=self.device) - bus_idx)
+            _, nearest_indices = torch.topk(distances, k=6, largest=False)  # 6 to exclude self
+            nearest_indices = nearest_indices[1:]  # Remove self
+            observations[i, obs_idx:obs_idx+5] = self.frequencies[nearest_indices]
+            obs_idx += 5
+            
+            # Renewable generation forecasts - next 3 time steps (3)
+            # Use actual forecast values for next 3 time steps, averaged across all renewable sources
+            if self.renewable_forecasts.shape[1] >= 3:
+                # Average forecast across all renewable sources for next 3 time steps
+                observations[i, obs_idx:obs_idx+3] = torch.mean(self.renewable_forecasts[:, :3], dim=0)
+            else:
+                # Fallback if not enough forecast steps
+                observations[i, obs_idx:obs_idx+3] = torch.mean(self.renewable_generation)
+            obs_idx += 3
+            
+            # Time features: hour of day, day of week (2)
+            observations[i, obs_idx] = self.current_hour / 24.0  # Normalized hour
+            observations[i, obs_idx+1] = self.current_day / 7.0  # Normalized day
+            
+            # Total: 1+1+1+1+5+3+2 = 14, need 1 more for 15
+            # Add own capacity utilization (1)
+            capacity_range = self.power_max[i] - self.power_min[i]
+            if capacity_range > 0:
+                utilization = (self.generator_outputs[i] - self.power_min[i]) / capacity_range
+            else:
+                utilization = 0.0
+            observations[i, 14] = utilization
+        
+        return observations
+    
+    def _check_termination(self):
+        """Check if episode should be terminated or truncated."""
+        # Terminate if frequency violations are too severe
+        severe_violations = torch.sum((self.frequencies < 59.0) | (self.frequencies > 61.0))
+        terminated = severe_violations > 0
+        
+        # Truncate after maximum time steps (e.g., 1000 steps = ~16.7 hours)
+        truncated = self.time_step >= 1000
+        
+        return terminated, truncated
+    
+    def _update_time_features(self):
+        """Update time-based features (hour of day, day of week)."""
+        # Advance time by dt minutes
+        self.current_hour += self.dt / 60.0
+        if self.current_hour >= 24.0:
+            self.current_hour -= 24.0
+            self.current_day += 1.0
+            if self.current_day > 7.0:
+                self.current_day = 1.0
+    
+    def _update_renewable_forecasts(self):
+        """Update renewable generation forecasts for next 5-15 minutes."""
+        # Simple forecast model: current + trend + noise
+        for i in range(14):
+            current_gen = self.renewable_generation[i]
+            
+            # Add trend (seasonal pattern)
+            trend = 10.0 * torch.sin(2 * math.pi * self.current_hour / 24.0)  # Daily pattern
+            
+            # Add noise
+            noise = torch.randn(self.forecast_horizon, device=self.device) * 20.0
+            
+            # Generate forecast
+            for t in range(self.forecast_horizon):
+                forecast = current_gen + trend * (t + 1) + noise[t]
+                self.renewable_forecasts[i, t] = torch.clamp(forecast, 0.0, 1000.0)
+    
+    def get_full_state(self):
+        """Get the complete 140-dimensional state vector for centralized critic."""
+        # State components: frequencies (68) + generator outputs (20) + renewables (14) + loads (30) + time features (8) = 140
+        state = torch.zeros(self.state_dim, device=self.device)
+        
+        idx = 0
+        # Bus frequencies (68)
+        state[idx:idx+68] = self.frequencies
+        idx += 68
+        
+        # Generator outputs (20)
+        state[idx:idx+20] = self.generator_outputs
+        idx += 20
+        
+        # Renewable generation (14)
+        state[idx:idx+14] = self.renewable_generation
+        idx += 14
+        
+        # Loads (first 30 buses - most critical/largest loads)
+        state[idx:idx+30] = self.loads[:30]
+        idx += 30
+        
+        # Time features (8): hour, day, hour_sin, hour_cos, day_sin, day_cos, load_pattern, renewable_pattern
+        state[idx] = self.current_hour / 24.0  # Normalized hour
+        state[idx+1] = self.current_day / 7.0  # Normalized day
+        state[idx+2] = torch.sin(2 * math.pi * self.current_hour / 24.0)  # Hour sine
+        state[idx+3] = torch.cos(2 * math.pi * self.current_hour / 24.0)  # Hour cosine
+        state[idx+4] = torch.sin(2 * math.pi * self.current_day / 7.0)   # Day sine
+        state[idx+5] = torch.cos(2 * math.pi * self.current_day / 7.0)   # Day cosine
+        state[idx+6] = torch.mean(self.loads)  # Average load pattern
+        state[idx+7] = torch.mean(self.renewable_generation)  # Average renewable pattern
+        
+        return state
+    
+    def _get_info(self):
+        """Get additional information about the environment state."""
+        system_freq_deviation = torch.mean(self.frequencies - self.nominal_frequency).item()
+        
+        return {
+            'time_step': self.time_step,
+            'mean_frequency': torch.mean(self.frequencies).item(),
+            'frequency_std': torch.std(self.frequencies).item(),
+            'system_freq_deviation': system_freq_deviation,
+            'total_generation': torch.sum(self.generator_outputs).item(),
+            'total_load': torch.sum(self.loads).item(),
+            'contingency_active': self.contingency_active,
+            'safety_violations': torch.sum((self.frequencies < self.frequency_bounds[0]) | 
+                                         (self.frequencies > self.frequency_bounds[1])).item(),
+            'current_hour': self.current_hour,
+            'current_day': self.current_day,
+            'agent_capacity_utilization': [(self.generator_outputs[i] - self.power_min[i]) / 
+                                         (self.power_max[i] - self.power_min[i]) 
+                                         for i in range(self.n_agents)]
+        }
+    
+    def render(self, mode='human'):
+        """Render the environment (optional)."""
+        if mode == 'human':
+            print(f"Step: {self.time_step}")
+            print(f"Mean frequency: {torch.mean(self.frequencies):.3f} Hz")
+            print(f"Frequency range: [{torch.min(self.frequencies):.3f}, {torch.max(self.frequencies):.3f}] Hz")
+            print(f"Total generation: {torch.sum(self.generator_outputs):.1f} MW")
+            print(f"Total load: {torch.sum(self.loads):.1f} MW")
+            print(f"Contingency active: {self.contingency_active}")
+            print("-" * 50)
+    
+    def close(self):
+        """Clean up resources."""
+        pass
diff --git a/requirements.txt b/requirements.txt
new file mode 100644
index 0000000..7b8b3f5
--- /dev/null
+++ b/requirements.txt
@@ -0,0 +1,6 @@
+torch>=2.0.0
+gymnasium>=0.29.0
+numpy>=1.24.0
+scipy>=1.10.0
+matplotlib>=3.7.0
+pandas>=2.0.0

From 666e701b7c9f326cd5615fd74110d38b50a5af25 Mon Sep 17 00:00:00 2001
From: Derek Smith <dkbsmith03@gmail.com>
Date: Sat, 1 Nov 2025 17:05:28 -0400
Subject: [PATCH 17/17] fix

---
 .../__pycache__/power_grid_env.cpython-38.pyc | Bin 9909 -> 13765 bytes
 main/power_grid_env.py                        |  34 +++++++++---------
 2 files changed, 17 insertions(+), 17 deletions(-)

diff --git a/main/__pycache__/power_grid_env.cpython-38.pyc b/main/__pycache__/power_grid_env.cpython-38.pyc
index 7a3e059720f1efa36c075aa4c959fb183f8228a1..a05f56f71c2faa8ae15616f81a4c1506f1341e95 100644
GIT binary patch
literal 13765
zcmb7LTZ|mXb?w*8^gMR<!6lb3QBCT_k;&ylqD0ZOxRgk}EUp!iisDh^!EmaVJImRb
zWpxk99nac=MMnx^!E%r|u`R>9Kt8O%FakJ1VjxHq<jX()@Mr(TaRMkrejLC*4gx#p
z-0pd5B_uoO>gsxRb=9qN&%IUq%EUy@z~8^L)Ai>ohVgHF82)qcuz)LAL7<GXp-g2p
z%(8{M-7r_}vTep`PT7&XR5^{i)5xs4Wp_1O&aURlxz&6*zgj34lCl%!2`O7F7oRm$
zN~P}@D(zb<lW*GPDP>(YN}2XQn1+$N)NX|Jqvx;r&Cq+Pb<5Y@v%0Rli_M#L-D<8f
zd8wtnm;HJZ_nI#$FZxy8tT(Uaa#xn?f#(ku4(hAzMm6*Uuj)N^`smem;H^os*Vyc>
zdbsSJJnmKbzTiymsCUY{S`9;A*HPoJuCA_m8nw@Or;dA<UOl#2Z<bK#w0EtFb{bVw
z8Yy+6U&=$V$35k*R-4Mxez4YR2ENyB*2ij`>NlIa9O6x0uvV?%eJ370epJ=*bwRz=
ztTw#E?D>-RZ?yep4V^pUq02t{hQVmH!?kvZq(-Z%f>Q2cwYD68lGnN#`1+<Fputlk
z?KN68v>InuL%bt*zQ!9$VmxZCW>{~wTm8|hEmxb@FeE?ylb60wmtm_nF*3>Uc#Zlh
zJ}h?^B4aF;%qT5i5tLGqbD>@fBj?3>5Jssh?X`v<+2@<<k^8wdzOLG+vCi;cHhBns
zgNSdGO$Z@`uWYO1zJp)NPvc50S$mA3Sjx|+?0cEFj3ra$RQ}FH*;NHKamOfU)d@f6
zJ1Cv=^HQg<GU4Z0UroMet0^^oCsi)`OifFe@+a}jFQrgo<~>X8P_uU&i0MwX3++v-
zIkg-28MR04#eIj`r}pDMtL{+;aNnu!RUYoU)IoI!_c?W+x*zx5>H+m2?t9cj>M-tm
zRY^UJ`#yD29Z^T`nC1OyUL8ZqJ#QI~Q9gi0@kmsVMZh(HCD8dyygWo)hvj%!;@bgM
zf_H6wwOI}7?NzVdgg7rjsOIlpz=+<RMet!f#=kuGzI9EaA6#09BMXZQy|jhzC24Q}
zXyL>EdH=tDcIoO%=OVkd)-GkDOtS)s^n=K4R^rtYSt^XCdLp`Bx!P(sRS@MUlod_!
ziKb$?g5=iLD{H!SHA*Z0X1(Uu+_*y$GM>g|;|hKOPza4TjZJgQFpbdKw7Z!tvujlB
zE<>klAxw2`gz2t>Fw;#Tbh~MUS!G@Wdd(Wj#@(4U=72qjr5QthF0@wiDwC9P`(?6&
zGAp*qy~B4Dq-;JZ%Q|s;g>A}CNZE;`Y_VUKwUc-E+bv32zLEVLv^%*?*-0rom6V;v
zJ?_KIHf5*O4)l6jO(2{>y&W8{_^m37H_gfj%yx~<og7)U6XkZPT?psG-79-2<S3t7
zGQ+(q`_!IyIJWzlgXeql21DlKnp?7l${bL8QS#nv9A8gy<PPHPhtxilJcx_q$=SLO
zZ@ypNydQ#oPxU{E{$BF;JaDpJZPYuy7cToBBkA+~;{C{ZeGMqyYlZhjOwV|W?bWN$
zCaon;<QB_fyy?S_oi2^lP4vQ0<v7JVeDXMS!jP``RH76<1=PRr@@P9MJmXyfmU}_y
zukpoDBW<Xpk(Z4xoHO1t3EHoF@ls;V=Q`eNr=FNUb;Nt@`24BYM_L#xv@_m|SaiT=
zEIQV@^s4vT$>Yb5AAuS>etfjn_>w&1y@Z8Xy;^V7!*veV;?Wa{K9F_mtzi`TAtUdO
zmoe(DUx8L$hT`M0_6|P;w0Ovaj(VupR@H|F<5JGVdUpM83d_7~L!sW~!UxpryBS0h
zh<bh@iQM0d%r3ld>Rp2ph7lu|QU4$*`uU{j!k2q>9$onG+u!2fkIyBM_s?C6Q+{$T
zskPWEY%JXU8k)a5i6CkEm;b2qcq%!O-Dus4oaOqp<tP;_SJ(U~rNZ?!|E|R<MFfkb
z2awgqFhy>qQimj0Dp3l0y0$Fc)>9~?ry0yJ*uj9Htal=a@*-gswi!g}5Cov5qa0N&
zKomj~Wre>~AVd9V^>#E}YXxDY*7BE@>fi?;CeidQ(4w)VW()PIjY?nI%DJo6z^_P~
zkpqcQ(R3Ze0?Q9on?Xp96YY`jh*xBV%&8K#)>@6$we={ABB5UaSBWy96O3q%McIC;
zbvv?}Yf)wik{Ld7GAdMsVpN12L^c>)d3uO}B<XqOH>&FuidYcM4Ch~MFD+qIr$GGt
zS{0wW+|u<<s~M#_zHSB44h>3lt9rHJS9(Q)C<XH6hb1?10>82J$%8NR^S&8nE0tQK
z8U&Tf`^JBN_Sh@nP{A?fYyWxxp0j-HLaWv$@eGdDTFO6m{-qPAP9O6Fgh!#(R{UCc
zY_(c%9*bv6#tRyHer;VJ!^rH#6}*7Jn9W%QGd=j(o_WMR!V5p9@3mdtZO60>3rf-X
zt($U8+>M;XnU^+q*`1xkG-19cb~RvX|Bvg8YV~Pc!MmbXSEv#I(Q?_!6kv8jg1Ysl
zdBJ%7uQtrkSV<Gc0o>+0ma?HboH=7Nvju=(aiCPpCF`sq%EeOFZsWS44~^AHp-wuk
zvtf0ue7~l$<_xLbPh(2bV!!p2vB&6IkAn~3J(3!iifa$&-Z&KZU1jzf(*Lh**rBoM
zDwkiRvRn3McFXvEYs1`N&z<YG{>4~dp`KN4o4z_D{X}hf{`fvRUDO!q8{d@l1ii^E
z#p>7;#hO=nD8WMKiH}!&y#?pdpr|UyA(kc}cC{Mn`t5mr52p1l=UeY%u%E#L3<z9z
zY1L`wmP~yFw*=RIer~Z;(8rm1k%2%Hu{D|=yrL4n0xOC7L_fi7rylxP@Wc!7OW;(i
z0fsNAhN`rtNDonSZ4GR+l+w?!mRT<shuXbaZA4}zGFS9x`Em(fWbD%j*u&s{1ctNQ
zE?N)TMRUJ7i{CEOvuDiRri<q({C1hU?aqUv@=w&=4y_hTHgUjLz$&pn0I}K++qzjJ
z+wA>i5a~0}+ZeX4sm$wHeHg+7f!{L8L|5!hh=Ifhv6vBod#+@ZMc%mL&O&G_*)G=X
zN){rJLs~zd%jt2f><V-8$SDkJRwfiU*9z_C)K{^@6l+bucFxEXMS4PIl*uQ=ix6vj
z!$CWfc<R3aoY>xNpbxh2y(#2REA$47nmu7J$?RrWPq7EmZ&i?9??`$zXw9h13ZvYq
zAim&>XsI%%iU@ZrytlFktZwg?@rQ;OVl=JXkj!ttnwgU&`6lauIc^>x_rv|(En@?K
zv+3cU?xr>mZW)`0050hzdIffLe$wX|FB8axv<;g(VG9gLW*e*onrItV__ECK7%?1D
zEU<acYh4w#BRbtX+%sy9&_t$A8!#8LN;)6Tcr<Nk!|L_ekx*3%9T9U~yc1YAc<T3m
zQP$|SFV6n(*?;ZaIj5-y^c;hG5!{_XkSxpRKj^L1qs*4TL+qI(|L}sW>JfOS&gCj6
zod`%L$iQDOY?yBwo5q&$4eN$;IbLJ1&r`KVtLc}l$eKT{e;0+{H_Q2C7%TYV!CH)R
zXSuR#t<|+B|AtNT9chT{*pd@1JAYQ(3c-_z^eX8~s8G63Kqkt=9nuE<;1bkNA_L10
z5Yk%JzsCm$oeBmZq)#(3-{UIqdI$!oW(-!OP|<YX4N^(mATgH`luD5j2l{0eoE(sY
z*l?X4BqVKMvdYDqtwsm{s{rd7c;U=2KPG><=|_{pX=p6UhG6%?SfWC$t<f;KNn1zx
z-jk}XNBb(*{E(vykBmr0P^qu3`l=48)i0rs`Y8s~&nVkoQ{+{HP7Wx|^hX*3V`1AO
zb)>CPs@Yy$TQ6k<cV%$(=b7$`kwEU4#6TSj!(9Th)3GL4`~`ML0Qp(O-oh0;iokGP
zGiMgeoV^!NJZlwA{Ola!yFo=}g|G;zv0d{32#jmGR%d1epkq#EMVXip2D*%9YZT$$
zAHSl{Ru+3~MfeIpNw~@epcv2s6Br>l#)buWPIs*>6IwrmG*>vl1?YM}Zx<dQ0F5Pg
z0$Pr-Zfdu&!Ls@OlZ#dga38pEG}v$~IA@I;r(^i;+C8vF3AP!7b`IqT@eE;ZZx$11
zn@sLg$$dJx&m{L9fIPIZIV+qbw*mO2F}}gJcWusf3!+?ThVLS5!YBtjy6(JdYM9t4
zu~*qFfm-`e3vY&s*{{HqpiuTB^&X@ayAy<3!fZF&&2{tL;!+C0>QsM>LG5GH#oK-j
zCaH%7=uwzsQx(=9&F>+&-3DDhYI7%(`U$97iUBiTN8)Bo>d;z!?dXXk-ic8wwcpYI
z1C}rPx5HQ$4Er}nt!)iEL8~*~<>jjO>7eDKc$TY;M&b_czmtfMB}&UF^ik>6Wgo+$
zJ%6npfd6}NIS;Z&Yw32g_VJZ^xg&LA-E#hubxFyJZOKnznFzwh3d4)J+)1XA#_4km
z9$`RRGs=mfRB6<&Cx|0vsL%XU47j|h9AIrm1z12~C7G9KXAh7B>X!T}NE6`ko?dbg
zw%{WU!g>v6<0`z}G>>$P?afwd%YN-T%}F?w=|mus&-H7r!_-@?sP$%bwO)hST7hS}
zCQf^xALk{SNm}TQGJsk)*<ZoRG)SfoyVY*^O@YwJUTdxCH5Rze;06N%iq;59lQAgX
zLcBZ?n`nu9yF7&t3+gl=lN9}ml<L8_lr9&QAJic>NuhFKwR#)aThplQ$4|<SpGYs|
z6LP)2gL`Jf-o%&t3W7?v%`V<#An+>z;TMz@#_$J-1&<*xinI0<@U{pPc0r4a)~sE$
zciCxTZU0vQ?iN6faR+?{0#A=1Fk2kq59Ggn!2Jpy0J8$OfSTC!gMSi`OOSCCysvY{
z69gJC&P^JO1lhDAr1B2exqu@1D14s%HZp{1X(NL+0B6!hCTSz1TpDSB(TvIgr1G8P
zuZSANa_)(jY;uSZ)K`tYtq_PsYzOK4(0`#wvMk~SA}s0<GY%m55SL*re0^cD<i?9e
z;dzLAEb(!jf%s;Y5i8juIpv}#(VJBb%}vfQ<ag*Nj!mKNqW%OBHz*)t*t^Ub=*C%u
zorg!}rN1irA3U_4J~Ck`w8b?Ld>a7<5oXGkxoH9GF=?>GU&P%BzA0{4nJ5gpOcN|&
zDFt^q^dznfZB><eLaglO3#_NV$%%uDokl?h=Fhu`WyN}irp2^UJaWo~nPYWLF)|<&
zNR#$pu%K;BWCjzN?HqiX7$68c?ojk^5-a9qmag=OFIl7JY&{dx=2AYMCGtA`I}CO*
zs52nHi)W5|mRe>`CXEo7Oqvdu6*3rITzUg(dpI*JZCJBr+JZ~P!g_)OCTHc$`_0b5
z(K*@%&qk)Ew}T;5^K-O1zLwxt;jpHJpd7%E!gd3hvBhD43F=~VDt36>*e@kBA5+pD
zD*24iq3mf{rDA91L2fwhBoq@}XUhTsGKKP_yRlQJDC8+WYwQymgteXPQY0qf2I<+R
zIkYf^G}?>P$qJwDukf5oV}diCBfXU>aPIZs+>5uPf?M2_!tSN4m}HecfRWKO@dvAP
zu{0CU^F<~P+!0wHgqGW#zP`d7QCYNxG#6$(<u%5{WJtsE!nUve98=CPp!SPbGwmZR
zdHN&DlY-}fZo2HKREYLbhMtSMx*g5-ERXdH{n;^RFHiXml-6}<sn~7WZ)Wvs6<fkJ
zzdTY=*U9#8`r~r<CA{kh<!(13hCM^4hFrOH0esp8I>51U|ET08RzWbTGe2JV{vJ(Z
z+IV4wZeY#I78*92b{JcxaApXivfs4yQ<4Lr#GQbRZMIpI*|3|Aa=@s2<r<15h$FJj
z*%Hg4Ek{bSbgu;8n^oyO@CqUpKV|UbJjIqNgm8}tE2nz`e#LHT&d5^~Gn=_BLy+US
zr+aij$S4f-A;w1MwE}DHi<WmXrk4I~Fc<7w4E*8q`fUt`c)xUqF{%N9n1T3Fq738h
ze_jCG^u<X8GO~j<9A2O}I3A!i^>AcYp|qu`$MG0D@(E3ec<HYpUY;HZ5m-vK5w4dL
z7e77rK{=Or=@E~ngX)qWu2*i>TR5`-fegx%tUi>W@9<S)5cXxHU7{%QlMK6H&e%C~
zujL@zkI;ddDgHKXx%Q0JIWQ_v<6LV*l+dK?!U~Q(iic1HNsJL}$-@{)=?5{f!V~zM
zhZ4CgWfTzXG6V}#=yc9K3&)MP^*LEX2C(P@e{fvk@X4cr51y#J%TJ%ba6Yb6idBQm
z5*&D%<V2XCkUeY$aV?Jz3EAHYhlHGlsQ%GUAGgLwK0S|G8C>$||Bi-+AsnBKqc88+
z^o|Spk@d2hGHr!i72-XA*L?ji0qe9AZP=cUBK#8ENs<8eU7TTn>n^2WRZ325k#3ih
zXA9fZf!-Z&hj5R+!WvwGTTc23n^*dlM*1YyoWlb+@7Pc~nGLt=Dw{4v^pHJ=-P?N;
zdgjg<U$v56x=AmmpajqxmA#&VCHex|<X26>6^T|5W?2LImTa6$|Ikvl*HUwTGG;T_
z#{WnzOD6Bxq4N8Nzu8!sRfS%P0$wB@_?<E~u!ngE#$~%%Wr>yiQ}inyxqfap3q_X$
zPEBm&U=8E_s>pV3oa~wq)yd9jifZ3214m%1c{GvS(kF>-qSClxgvi{b+&u#in|VFT
zG^!ehXTnAC_T9CS10S3XuVTzXjnloY$NdX2Dv_&}=7uz(*lC|(<76q3v+7ryLkK0_
zBT<Y(endNdodwg?*lABq8adcQxUQdJD-Sae)!ATdmBA?nT;%{iSyHq|qMeDZ5gh8P
zLD7`@k5Hg2yM9!cSp}aT^<N+^#qwkm`;Yi+=vRf!;GkTrRdKip+sestkbFDDE^|ka
zZhZaHyt~2mdlnG`Y;4Awg1zHFLjuN<tAIN<1q$X~({*O7lXzn8E@u{H0f4ZLI)_IA
zZ+z7cd~hS{pKi_+t|91y>z}(wrsy*4btPI1QmpLj4$T@EO3w8Z*1XlV#eJRXJQDNX
z`cmIp5Wq9oP~KHITKYZ(?vVG5!uePWz0o_8^Ty(WU|;<|AUP!`mWp@Kw^#tKBZgGY
zSa~`Cnb!WWzW6E&{sDtArSS&Re!vlS5i#J)bM2hfxp#E5Vy4^cC^XW@AW_;mP%=o1
zh=^qu<G;BN3>Nkwp8y8X6fi2lFA+bTq~i%2ic*S$XJPOh{Q^ytilykJN_QShSS7s<
zM`LB*^F4Tn>0;#x1vpoaFkt+a?_Y<)5LWqT=uvD1{yAb)9U|<)1x4b&!2)B$@<pWm
zEr%sG)F?=FaOTX;q0ym9sMv@<WMp)xKKd%;T|9(1Hz_>+E3~7bHBEpn1c`J?+1OF)
zyE2utFQ!y>m|p3bXjzrQ1}rux$>~u)c5gl!(<cCDW)4J*G^z{hdk(uPVR|!1iWS?`
zfI@J80lO<V+{C^<n~d*vwltNq?6c$GA>^}3x7w+rCr<T_ugu3A6SCcSgd-{A&3#CH
zkijJeLbBdQj2n#tQ_FY&fK|g~a4I;Wt+jeQFUMJ4F}T2hKo?VXB393pO@==3n!uS<
zPKD^ag2%g65B$gs%WkjV<5Np-)IV`*c}#;eDQ6#m-=v7m#Br2JCbvH<g98O4Q%orD
zqeK}^53>aBmxFP`xWi;|zZrHi^_FbdiQr2PO)}<y{U$jMG$8AwlN@m3Bqx>RWK=rV
z*+~gZeNrNml)z52=t@gA2eS%#k7c>Hebxwbo7iy<b1SmV$1O^h7JY1~d{ScKV@nja
zX@f_~@Q$R7iEY{#Em4#b@Rj2`d7zJdn;c3@Cuvhm6Wu)Bxk%kihcrEm4Il36@l&$0
z1L)m^eFlN0(AMBu&{ZBJQ(jGDcdj<9sCm>>=NKGCuy|L5Riv1J5eu+L_Mh^R02lAq
z{V`*E5WtrxQhSjJ)Ksz5))?~{P}pPZRG>#t=X*@sWUz$*2P#{E{uW}1rvitExQ~hR
zG3xf11V=?`ucdYaJ7_HPGYa#AsA@nxI#y3D@=iD}g7+Uve%Lv;I{QWtY_zQrxwg>c
z1e?Vb`~?DVHf$uWSh%pA3?Gw&y#=PEdO2|#p3*EAKXu8EGkA9UxqQmTPl(aAz!bKd
zSAfIoM*j&MGS;ws7g7tjCS+R~S{B>NMLbUuPT#}^`FlJ+k?uTB3E6%d)~n(a-AA{N
zWgO=R7T`yQA&sw>JykLz*;73t9}c11!)^(R@0M<lrtBNpbG2i6U?W1b;J|;KapIXC
z=%E++SV7=m3PdxfE$cTk=FD%${e+x7ZWU~|=L!NxRuC0Q6Nk=vO$=<&AXNH2mYE7$
zAv_s_Ezl?zr_jk6``4(uVu+^vMtkG)9epx4#NtX*@p_;*57AQ^#c_{pzV{tI%JNuZ
zOni78;=wy&O+nI&ae9tf>}$LlWUyqMZXoSfTmtkHk|Toe;b%!08@(<#paU<S`y1z1
zon0g1-`_EfQp;`ZOOI?nasl*p_Yd$OCqOLyedusPv-ut4yD;)_4g@xC;tp}Pa31$N
zFs{C9Cpl^4h;KZ}$>5o$Vy~q8U3dVcjVy9R&!c1xZRGoH;M}p~V9|`_(8S9P^`pqP
z)Fc+dROisk(23Zh8r#0}4#&q{sts_YAL6q*>C4y%IOBDuE<sfK|6s6=6np-WrH+jg
zr=6K+`X^DvnmFUV*0GLw9p`n)y+Q#R^1(~p6Rdmb)lMPKr7$6V_UZl}(7MO1rJMLy
zhtA2P|L?!WQa(zpX`DTZw|E4F`%)L#+~KydEwi1hB_ah5g3Y^egiVu7_&UqwH-!N*
zEEi{t_L>jl+KX6c$A}QcrAI&m0*R9?2`P3EQM4?r&W@)W9O5?qqr?xuTwxK-FAj<H
zC=Zn*8i7B{eoz>V@hx<m;0S!MqJa*}qm(W%AarYz9NG0MS&R6e44LlIy*!hOjW!==
zQgf9GMxc$&u1Zv>RN$L!BsrN%MYU>3#u-ig+lHLq)Z{r}Y;b-ZW6M0FrMW~jl|+1)
zBhyq>9;DIFA}MmNwptDS7IV7{=xEnpVnBaYyg%1u><tFw-LhY(zt0%gw*C==zeW&E
zBquU!K6GDYXwsf%5kUhNu{m79QwUsGikOAL@7ES^fxjGnxc_?UlIuPA#Gy-W!QJ5=
zbnkI9?x)?nn{%h#L+;L8UbDT@wEjEX%Gv(EQ{bfXdVIh&Mmmk`<RlA@?83<%4+rO6
xMod$G#N#A-zslz9i_M$T{?LcW^z%a^_iQ{ZPg06R2%QP;<>4F!j=UkX{}%wFMKS;Y

delta 3567
zcmZ`*Uu+yl8K2qN+uQqZ-<|FApWt(pINl^UPU8>~lE%1cLRIS4Nt!fA!ew)_cJ6$)
zm&~5i+MKx}k|L1OR?Gwcinv2OP#%hcg%A(?d4M7wctd^I2PhI!ls78W2S6fx-|X2=
ztZ=9O?aVjd%zXR%{>=P->E|ybUrr>V1b$CnHOg<y+)I96OaBGB4k=Jh3L2*c#wX;j
zv?J7#A8W1Qy?htneVZ0?e4OvOO$viwBRVMzog>!1^`A0IX;j|Rzuw}pCMhO(95C6?
zDyb%Ik_zL|yBd#V2~P|V80l$}c5jDJ!jsTT^AupFk*)Oc9^MQ3p#|FLuMF_?T^&Yp
zI0F5_7(v8wcrpvxBRWH{;xMcjCOiWsW!JxL^lEfb{=}S!&dn_1hZ1s04o6O~Icr2Z
zkt_1Qk+ZZQ=cDuTqv)Gw#$kL29`6`xW}{l18e4ysr~4W9hnT6uo9R>|8Zn6e&<yDm
z-XyBp@`Knx`nojZ<MhY!Ks+a(k59=~ygqt+jWtL+%uO&IsOYfE+8SNcfgXK{319Xm
z`sKdFi#z-VsdTkK(_9GXql$=#{8nQ2^a$AZfC31#R}3S30wD(QK*tu2TJ(ViukNRR
zcyi8)iaj9kL#vLcdvZr|cYHtg!@_oX*_D@*BlCw*K+A=w&1>aG!*)xKPm4aSh)=^v
z07tPk32*@(Zx;a3b1bRtWJ#K%S@`$UJWIpdq&fMg<W=JtRLsglsTKJ_>Oc#8S}{2K
zFw8aeVQKj26gRpOjfb-sB^i)HG*vd`Aj=|w#Xb(S6#F?w#{hUf*CA`LUV9MUMl;kN
z>ZoXF8Vec(LuB=j(R}xJ*74-8dwz2jNxc81?0aYawtD;I246`Kist|V(q>vSr>&8I
zv{=n{m+flNa~v+FK*cwZwt7uGiygi0I-cmk`TYp{<a_BHJt6;)9!n^aL)E&?JuxFc
zPA|#dW%kFAX*E}zmjB7@|I9Xir(yUOJORI^NuxAQqihJ6&9W$9j`l+fL(z~TyE}gd
zluPLFjh_jP5Il<f9})SB?0BmQaSC*5M<8~K2F9aug~nNj0hK^>+r+elcclA{cHc3e
z7IfP24uQzk72OF%H?EoFx>^&`xt#)OxM{R|zzLA#P1e+#p=Nj?bdC`#v>xeypPs>}
z>5}}`Z|^IUXWp6-n0r=QF)l1<2V@?^)?S1(!ixZYs8qFUH{@JyREpfiLl}zUWrT|e
zDTG%5EK?|=uRyzyxKV%A5yjGV+g*e(s1J@F6JJ2h%LtD<z}LNobK{3F;9-`LWR~@?
zBug{1`<EXL{$mlNXz{z>!QZ<lRsN|^Qs1Eh>3%Ty%ag%3g1Y!3z#Q1M2Rm~JTVeAO
zc8dsEgtGt^6Nlx@u+{gtec~#N{h6$Pd3XnHA#LZoK^*K|2Fd9;f%aLkU@H}ddb#1$
zL?4L7IfNMmwf}i+sr}W~7{$0XN-rT45VqN1gX}{>)(3aEw1s;Gb!Z2+RDnXn<~AI9
z0>x@CT-=X1?sCl$cEb^$#SwI?m`A|R4orCgTl)Zfqh>d1%T@6-N<N98VnK10fvz<e
zc)E}ZoLqJ*PN`9@yJA+3j$Yo*-Zhx=Pa@wQJu8>2OskDmR)dG-NQbUb$PB2=oMy16
zVUOcz2qkccVI{{Ip$^TG0n}k!U(?mBU>zuLgJcQvNUdsc1A6)_YKAZyu)rN5Q2n(g
zg99<650DsEE~6dp5U5z@>bzXfPqii%gmZJ*aZ4+C?%XQd_(5JC>Q25?_ZprxDUKr*
zpF=>GiweRf^(xl7q2@FDnkTjhW!`cv;2h}_O8{0l@OK^Bg<jZNE{~;pQj3m@Zi20g
zxT}0;Z1=Q-bGAi831nH=+rt1NZ~!FPE;uLQ{m{_jEGZw3O}4Nqh2c?EX&SVl#&41j
zbm^SlW#9lj9kj2~Q{-)WCBBB_L)jXH62zhSKrh^6g9Hvwt*I|)$4OJ?Oc4r4=Kmdu
zG_}@~CV}6g-J=GsEy@J&i;9F0$P?f%orkz_7td@1YFn5y9=?O-VEu&t#;Rb3IY91c
zU6agBlQuNY>rYYv1E6qL!X3KxMIeJ)(Uy!upy0AM^G&MGwDB`eBVVnT?5h0A&XLxn
zO7<<^KQMsh)0XDbB|luX1ysDooEQM1)e{s1m5;q35C;%2DaA<ym5wS>R6S6Mc>@M~
z!-g|&(NRh4>y7%2C10;Ow(B#mufaEJttzRrxR5L7#&eg@)rFMn*aA!_mf@~Zg0pg4
z@L&O-z~GsH2GRPp9vY465JcwYGXgh?I2{5BK3O^>|GjfkPVD*K=T#}a3G;&Z#&f{Z
zRV){JJ?EAqoZ_u=y&9yCSU|-^gl#r(kX?|eiP=^hS}H<aXn+?;gSJ(9=+L&_!5+pZ
z+w7?u%40n<=z&%Qv~e5`&q|cjZdnL=5JEv&z*RYWCk&p|SC7cQPM+9{wG%QDKgLfY
z%f-!N;3qblMX$kS&(vs3m2-h}V!=VENO=te0V$Z^Vg*4FrJSnVt0rO%c$Wt0zl{|<
zQ0=X>)@ic)4^rFw*J}Su^;ds(Fh=3t8-MEki)e8WVVk@AtX767soAeRX?IIg8$&GB
zxwFkFaLsaDkh)Y_hZe__Z7tm3aVzmYQ1_e_D(@Y7&UhUZtl@7?pK9T)Ufop%79GJ5
zrU3k~%WWa-6)}xH{aU?V71wd}DuRuGXM`vrU~meIj^K822U~9d_=#c>?jfF6a;nv0
z@e!Log(^o$0uU=6rk!_8{(XA<)NwOEzJKRaX57q}V`kQzG-GBW8dGXkulNRN79xu)
zHP`ma%e4Z7rqFOT<<D2JH6<?`-Zc(%*^TSsd7QwWb#JK^9x*`?bX5M}@c91#6RqcF

diff --git a/main/power_grid_env.py b/main/power_grid_env.py
index 2aeb371..ee63125 100644
--- a/main/power_grid_env.py
+++ b/main/power_grid_env.py
@@ -109,25 +109,25 @@ def _initialize_grid_topology(self):
         # Create a connected graph structure
         for i in range(self.n_buses):
             # Self admittance (diagonal elements)
-            self.admittance_matrix[i, i] = torch.rand(1, device=self.device) * 10.0 + 5.0
+            self.admittance_matrix[i, i] = (torch.rand(1, device=self.device) * 10.0 + 5.0).item()
             
             # Connect to nearby buses (simplified ring + radial structure)
             if i < self.n_buses - 1:
-                admittance = torch.rand(1, device=self.device) * 2.0 + 1.0
-                self.admittance_matrix[i, i+1] = -admittance
-                self.admittance_matrix[i+1, i] = -admittance
-                self.admittance_matrix[i, i] += admittance
-                self.admittance_matrix[i+1, i+1] += admittance
+                admittance_val = (torch.rand(1, device=self.device) * 2.0 + 1.0).item()
+                self.admittance_matrix[i, i+1] = -admittance_val
+                self.admittance_matrix[i+1, i] = -admittance_val
+                self.admittance_matrix[i, i] += admittance_val
+                self.admittance_matrix[i+1, i+1] += admittance_val
         
         # Add some additional connections for robustness
         for _ in range(self.n_buses // 4):
-            i, j = torch.randint(0, self.n_buses, (2,))
+            i, j = torch.randint(0, self.n_buses, (2,)).tolist()
             if i != j:
-                admittance = torch.rand(1, device=self.device) * 1.0 + 0.5
-                self.admittance_matrix[i, j] = -admittance
-                self.admittance_matrix[j, i] = -admittance
-                self.admittance_matrix[i, i] += admittance
-                self.admittance_matrix[j, j] += admittance
+                admittance_val = (torch.rand(1, device=self.device) * 1.0 + 0.5).item()
+                self.admittance_matrix[i, j] = -admittance_val
+                self.admittance_matrix[j, i] = -admittance_val
+                self.admittance_matrix[i, i] += admittance_val
+                self.admittance_matrix[j, j] += admittance_val
         
         # Agent-to-bus mapping (which buses have controllable agents)
         self.agent_bus_mapping = torch.randint(0, self.n_buses, (self.n_agents,), device=self.device)
@@ -453,7 +453,7 @@ def _update_renewable_forecasts(self):
             current_gen = self.renewable_generation[i]
             
             # Add trend (seasonal pattern)
-            trend = 10.0 * torch.sin(2 * math.pi * self.current_hour / 24.0)  # Daily pattern
+            trend = 10.0 * math.sin(2 * math.pi * self.current_hour / 24.0)  # Daily pattern
             
             # Add noise
             noise = torch.randn(self.forecast_horizon, device=self.device) * 20.0
@@ -488,10 +488,10 @@ def get_full_state(self):
         # Time features (8): hour, day, hour_sin, hour_cos, day_sin, day_cos, load_pattern, renewable_pattern
         state[idx] = self.current_hour / 24.0  # Normalized hour
         state[idx+1] = self.current_day / 7.0  # Normalized day
-        state[idx+2] = torch.sin(2 * math.pi * self.current_hour / 24.0)  # Hour sine
-        state[idx+3] = torch.cos(2 * math.pi * self.current_hour / 24.0)  # Hour cosine
-        state[idx+4] = torch.sin(2 * math.pi * self.current_day / 7.0)   # Day sine
-        state[idx+5] = torch.cos(2 * math.pi * self.current_day / 7.0)   # Day cosine
+        state[idx+2] = math.sin(2 * math.pi * self.current_hour / 24.0)  # Hour sine
+        state[idx+3] = math.cos(2 * math.pi * self.current_hour / 24.0)  # Hour cosine
+        state[idx+4] = math.sin(2 * math.pi * self.current_day / 7.0)   # Day sine
+        state[idx+5] = math.cos(2 * math.pi * self.current_day / 7.0)   # Day cosine
         state[idx+6] = torch.mean(self.loads)  # Average load pattern
         state[idx+7] = torch.mean(self.renewable_generation)  # Average renewable pattern