docs: switch install to extras syntax for tracebloc_package 0.6.33

join-use-case/start-training.mdx

@@ -1,6 +1,6 @@
 ---
 title: "Start Training"
 description: "Step-by-step guide to training a model on the tracebloc platform."
 ---
 
 You have joined a use case and accepted the terms. Training a model is easy: you run a Jupyter notebook locally to connect and submit models to the workspace.
@@ -10,7 +10,7 @@
 
 ## Pull Training Notebook and Model Repositories
 
 On your machine, create a tracebloc folder and pull the [Training GitHub repository](https://github.com/tracebloc/start-training/tree/main) and the [Model Zoo GitHub repository](https://github.com/tracebloc/model-zoo/tree/master). The notebook contains all commands to connect and start training, the model zoo contains a selection of compatible models ready for training. The easiest way to customize models is by starting from the model zoo. Open a terminal and run the following commands:
 
 ```bash
 mkdir tracebloc && cd tracebloc
@@ -25,7 +25,7 @@
 
 ## Create a Virtual Environment
 
 Create a new environment, name it for example "tracebloc":
 
 ```bash
 conda create -n tracebloc python=3.9
@@ -36,9 +36,15 @@
 
 ```bash
 python -m pip install --upgrade pip
-pip install tracebloc_package
+
+# Pick the extra that matches your ML framework:
+pip install "tracebloc_package[pytorch]>=0.6.33"      # most common
+# pip install "tracebloc_package[tensorflow]>=0.6.33" # TensorFlow
+# pip install "tracebloc_package[all]>=0.6.33"        # everything
 ```
 
+See the [tracebloc Package page](/tools-help/tracebloc-package) for all available extras.
+
 ## Install and Launch Jupyter Notebook
 
 Install Jupyter into your environments:
@@ -53,9 +59,9 @@
 jupyter notebook notebooks/traceblocTrainingGuide.ipynb
 ```
 
 ## 1. Connect to the tracebloc workspace
 
 Follow the instructions in the notebook to authenticate. Have your tracebloc user credentials ready:
 
 ![Log in](/images/Connect-to-the-tracebloc-Client.avif)
 
@@ -79,7 +85,7 @@
 | Image     | Classification         | `image_size` has to match image x/y-dimensions <br></br> `output_classes` has to match # of image classes                                                             |
 | Image     | Object Detection       | `image_size` has to match image x/y-dimensions <br></br> `output_classes` has to match # of object types                                                              |
 | Image     | Semantic Segmentation  | `image_size` has to match image x/y-dimensions <br></br> `output_classes` has to match # of object classes                                                            |
 | Image     | Keypoint Detection     | `image_size` has to match image x/y-dimensions <br></br> `output_classes` has to match # of object classes <br></br> `num_feature_points` has to match # of keypoints |
 | Tabular   | Tabular Classification | `output_classes` has to match # of classes <br></br> `num_feature_points` has to match # of features                                                                  |
 | Text      | Text Classification    | `input_shape`<br></br> `sequence_length` <br></br> `output_classes`                                                                                                   |
 
@@ -166,7 +172,7 @@
 
 ## 4. Set Training Plan
 
 Set the experiment name and configure hyperparameters.
 
 ```python
 # Set experiment name
@@ -180,9 +186,9 @@
 trainingObject.getTrainingPlan()
 ```
 
 Get the training plan to check settings before you start the training. For a detailed list of all hyperparameter options, see [Hyperparameters](/join-use-case/hyperparameters).
 
 For classical, non federated and non gradient descent-based machine learning algorithms like random forests, XGBoost, SVMs, logistic regression, use simplified settings:
 
 ```python
 # Set training parameters
@@ -217,9 +223,9 @@
 
 **Note:** The compute budget is set by the use case owner. Your team shares one compute budget with no per-person limit.
 
 ## Submit an Experiment to the Leaderboard
 
 Once training is complete, submit your best model to the leaderboard for evaluation on the test dataset. For the full submission flow and leaderboard details, see the [Evaluate Model guide](/join-use-case/model-evaluation#submit-to-leaderboard).
 
 ## Inviting Others to Your Team
 

tools-help/tracebloc-package.mdx

@@ -1,21 +1,31 @@
 ---
 title: "tracebloc Package"
 description: "Python library for uploading models, linking them with datasets, configuring training parameters, and launching training runs on the tracebloc platform."
 ---
 
 # tracebloc Package
 
 The `tracebloc_package` is a Python library for uploading models, linking them with datasets, configuring training parameters, and launching training runs on the tracebloc platform.
 
 ## Installation
 
-```bash
-pip install tracebloc_package>=0.6.32
-```
+ML framework dependencies are optional — pick the extra that matches your setup. The default install is ~140 MB (up from ~8 GB in older versions).
+
+| You use | Install command |
+|---|---|
+| PyTorch (most common) | `pip install "tracebloc_package[pytorch]>=0.6.33"` |
+| TensorFlow | `pip install "tracebloc_package[tensorflow]>=0.6.33"` |
+| scikit-learn only | `pip install "tracebloc_package[sklearn]>=0.6.33"` |
+| Hugging Face LoRA / PEFT (requires PyTorch) | `pip install "tracebloc_package[pytorch,huggingface]>=0.6.33"` |
+| XGBoost / CatBoost / LightGBM | `pip install "tracebloc_package[boosting]>=0.6.33"` |
+| Survival analysis | `pip install "tracebloc_package[survival]>=0.6.33"` |
+| Everything | `pip install "tracebloc_package[all]>=0.6.33"` |
+
+Upgrading from 0.6.32 or earlier? See the [migration guide](https://github.com/tracebloc/tracebloc-py-package/blob/develop/MIGRATION.md).
 
 ## Key Features
 
 - Upload model files and pretrained weights
 - Link models with datasets from your use cases
 - Configure training parameters (epochs, optimizer, learning rate, augmentation, callbacks)
 - Review training plans before starting
@@ -58,15 +68,15 @@
 
 ## Model Zoo
 
 Use a ready-made model from the [tracebloc model zoo](https://github.com/tracebloc/model-zoo) or bring your own. Supported tasks include image classification, object detection, text classification, tabular classification/regression, time series forecasting, semantic segmentation, keypoint detection, and time-to-event prediction.
 
 ## Google Colab Quickstart
 
 The fastest way to get started is our [Google Colab notebook](https://colab.research.google.com/drive/1N00idtpoaq1lk9OJE6g4bMqd8o-Qex2C) — runs entirely in your browser, no local setup needed.
 
 ## Next Steps
 
 - [Start training](/join-use-case/start-training) — detailed walkthrough of the training workflow
 - [Hyperparameters](/join-use-case/hyperparameters) — full reference for all training configuration options
 - [FAQs](/tools-help/faqs)
 - [Key terms](/tools-help/key-terms)