[Kernel] Initial commit containing new Triton kernels for multi lora serving. #5025
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Outdated suggestions cannot be applied.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.
SGMV Triton Kernels
New Triton kernels for multi lora computation. These (should) handle any shape and data type, apply to loras in a paged format and compute at the actual lora rank and also speed up for grouped lora requests (especially prefill).
The PR contains the kernels, tests for the kernels and benchmarks. A follow up PR will work on adding the ability to use these in vLLM.
ping @Yard1