feat: add ROCm and CUDA Dockerfiles with entrypoint and setup scripts

tazlin commented 1 month ago

Adds support for ROCm and CUDA Dockerfiles which support different python or driver versions.

tazlin commented 1 month ago

@CodiumAI-Agent /describe

CodiumAI-Agent commented 1 month ago

Title

feat: add ROCm and CUDA Dockerfiles with entrypoint and setup scripts

Adds support for ROCm and CUDA Dockerfiles which support different python or driver versions.

enhancement, documentation

Added new entrypoint script to handle environment setup and execution for both CUDA and ROCm environments.
Introduced multi-stage Dockerfiles for CUDA and ROCm, supporting customizable build arguments and pip caching.
Documented the usage of new Dockerfiles, including build and run instructions, configuration options, and customization possibilities.
Updated advanced README to include options for mounting configuration files and model directories.

Relevant files

Enhancement

entrypoint.sh `Add entrypoint script for GPU-specific setup and execution` Dockerfiles/entrypoint.sh Added script to source environment variables and set up GPU-specific configurations. Implemented logic to determine GPU type and install dependencies accordingly. Added logic to run setup scripts and worker based on configuration file presence.	+51/-0
setup_rocm.sh `Add ROCm setup script for AMD-specific configurations` Dockerfiles/setup_rocm.sh Added script to uninstall NVIDIA packages in ROCm environment. Included script to install AMD-specific components.	+5/-0
Dockerfile.cuda `Create multi-stage Dockerfile for CUDA environment` Dockerfiles/Dockerfile.cuda Introduced multi-stage Dockerfile for CUDA environment. Added support for customizable build arguments. Implemented caching for pip installations.	+60/-0
Dockerfile.rocm `Create multi-stage Dockerfile for ROCm environment` Dockerfiles/Dockerfile.rocm Introduced multi-stage Dockerfile for ROCm environment. Added support for customizable build arguments. Implemented caching for pip installations and AMD-specific setup.	+67/-0

Documentation

README.md `Document usage of CUDA and ROCm Dockerfiles` Dockerfiles/README.md Added documentation for using Dockerfiles with CUDA and ROCm. Included instructions for building and running Docker images. Provided configuration and customization options.	+127/-0
README_advanced.md `Update advanced README with configuration options` README_advanced.md Updated instructions for setting Docker worker settings. Added options for mounting configuration and model directories.	+11/-1

💡 PR-Agent usage: Comment /help "your question" on any pull request to receive relevant information