An InfiniTensor-Featured Comprehensive Accelerator Evaluation Framework
A unified, modular testing framework for comprehensive performance evaluation of accelerator hardware and software stacks.
InfiniMetrics provides standardized interfaces for benchmarking across multiple layers:
- Hardware-Level: GPU memory bandwidth, cache performance, compute capabilities
- Operator-Level: Individual operation performance (FLOPS, latency)
- Inference-Level: End-to-end model inference throughput and latency
- Communication-Level: NCCL collective operations and inter-GPU communication
- Unified Adapter Interface - Consistent API across all test types and frameworks
- Extensible Architecture - Easy to add new test types, frameworks, and metrics
- Comprehensive Metrics - Scalar values, time-series data, custom measurements
- Framework Agnostic - Support for InfiniLM, vLLM, InfiniCore, and more
- Production Ready - Robust error handling, logging, and result aggregation
For detailed guides, configuration, and examples, see the full documentation.
- Installation Guide - Prerequisites and dependencies
- Configuration - Input format and parameters
Contributions are welcome! Please see our Contributing Guide for details.
- Fork the repository
- Create your feature branch (
git checkout -b feature/AmazingFeature) - Commit your changes (
git commit -m 'Add some AmazingFeature') - Push to the branch (
git push origin feature/AmazingFeature) - Open a Pull Request
This project is licensed under the MIT License - see the LICENSE file for details.
Built with ❤️ by the InfiniTensor Team