Skip to content

Latest commit

 

History

History
31 lines (19 loc) · 1.43 KB

README.md

File metadata and controls

31 lines (19 loc) · 1.43 KB

AI-Infra Benchmarks

Welcome to the AI-Infra Benchmarks repository. This repository serves as a centralized location for benchmarks related to Large Language Model (LLM) inference and other performance evaluations. As the repository evolves, we will incorporate additional benchmarks for fine-tuning and inference to gauge performance across different hardware configurations.

Directory Structure

  • /nvidia/: Contains benchmarks specific to Nvidia hardware.
  • /common/: Includes benchmarks that are not specific to any hardware, such as serving benchmarks.
  • /amd/: Will contain benchmarks for running on AMD hardware.

How to Run Benchmarks

  1. Clone this repository:

    git clone https://github.com/runpod/ai-infra-benchmarks.git
    cd ai-infra-benchmarks
    
    
  2. Follow instructions in the respective hardware directory to run the benchmarks.

    • For Nvidia benchmarks, navigate to the /nvidia/ directory and follow the provided instructions.
    • For common benchmarks, check the /common/ directory.
    • For AMD benchmarks, follow the instructions in the /amd/ directory.

Notes

  1. This repository currently contains inference benchmarks only. As our requirements evolve, we will actively update the repository to include additional benchmarks and tools.

Contributing

Please follow the standard pull request process and include relevant details about the benchmarks or improvements you are adding.