Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Background
KernelBench currently focuses primarily on NVIDIA GPU workflows. This PR extends support to AMD Radeon GPUs, enabling broader hardware compatibility and evaluation capabilities.
Changes
This PR introduces the following updates:
GPU Specifications & Prompts: Added AMD Radeon GPU specifications and updated prompt content in gpu_specs.py and prompts.toml
Prompt Construction & Evaluation: Enhanced prompt construction and evaluation logic in prompt_constructor_toml.py, eval.py, and utils.py to accommodate AMD-specific workflows
Dependencies: Updated requirements.txt and setup.py to include necessary AMD-related dependencies
Documentation: Updated README.md with AMD GPU usage instructions and examples
Scope
Affected files:
src/kernelbench/prompts/*
src/kernelbench/eval.py
src/kernelbench/utils.py
gpu_specs.py, prompts.toml
requirements.txt, setup.py
README.md
Compatibility: Fully backward compatible—existing NVIDIA workflows remain unchanged. This PR only adds AMD-specific support.