Implement custom kernel for LLaMA rotary embedding #14
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
WoosukKwon
changed the title
bigPYJ1151
added a commit
to bigPYJ1151/vllm
that referenced
this pull request
Sep 12, 2023
hongxiayang
pushed a commit
to hongxiayang/vllm
that referenced
this pull request
Feb 13, 2024
luo-cheng2021
pushed a commit
to luo-cheng2021/vllm
that referenced
this pull request
Mar 25, 2024
mzusman
pushed a commit
to mzusman/vllm
that referenced
this pull request
May 6, 2024
* remove JambaConfig and use official one from transformers * changes in Jamba modeling file to align with official HF format
fxmarty
pushed a commit
to fxmarty/vllm-public
that referenced
this pull request
May 31, 2024
enable fused topK_softmax kernel for hip path
ykim362
pushed a commit
to ykim362/vllm
that referenced
this pull request
Jun 17, 2024
0612 kernel of FP8 on A100
yukavio
pushed a commit
to yukavio/vllm
that referenced
this pull request
Jul 3, 2024
Summary: Add benchmarking scripts and utils. Things to note : - All files are stored in `neuralmagic` folder. - neuralmagic/benchmarks/scripts/* : Actual benchmarking scripts that interact with vllm engine. - neuralmagic/benchmarks/configs/* : JSON config files that define what benchmark commands to run. - neuralmagic/benchmarks/run_*.py : Scripts that consume some config file and run the benchmark scripts. - neuralmagic/tools : Add tools Testing: Local testing --------- Co-authored-by: Varun Sundar Rabindranath <[email protected]> Co-authored-by: rsnm2 <[email protected]>
Closed
1 task
wuhuikx
pushed a commit
to wuhuikx/vllm
that referenced
this pull request
Mar 27, 2025
a fix follow up [MRotaryEmbedding change](vllm-project@bf3b79e#diff-6bc44986c91bf0876240dec03d56c748403691c7fcd90f7a22e7affff7b033ecR839) Signed-off-by: z00897138 <[email protected]> Co-authored-by: z00897138 <[email protected]>
1 task
1 task
1 task
Pradyun92
pushed a commit
to Pradyun92/vllm
that referenced
this pull request
Sep 29, 2025
ISSUE: The USE_CUTLASS_MOE environment variable support (CLAUDE.md entry vllm-project#14) was lost during a previous merge, removing critical debugging/compatibility control. ROOT CAUSE: Upstream changes overwrote the Mantle modification that added environment variable control for CUTLASS MoE implementations. SOLUTION: Restored the missing environment variable logic: - Added `import os` to imports - Restored `default_use_cutlass` calculation with original conditions - Restored `USE_CUTLASS_MOE` environment variable with smart defaults: * USE_CUTLASS_MOE=1 forces CUTLASS MoE on (default when conditions met) * USE_CUTLASS_MOE=0 disables CUTLASS MoE, fallback to other implementations - Maintains backward compatibility with automatic detection CODE CHANGES: - File: `vllm/model_executor/layers/quantization/compressed_tensors/compressed_tensors_moe.py` - Lines: 5 (import), 547-556 (environment variable logic) - Annotation: Added comprehensive Mantle modification comments for future merge guidance TESTING: Verified import functionality and environment variable integration. This fix enables debugging and compatibility control for CUTLASS MoE implementations as documented in CLAUDE.md registry entry vllm-project#14. Signed-off-by: Pradyun Ramadorai <[email protected]>
heheda12345
pushed a commit
to heheda12345/vllm
that referenced
this pull request
Sep 29, 2025
setup sparse attention backend
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.
This PR implements a custom CUDA kernel for rotary embedding, which is used in LLaMA. The kernel is responsible for the entire process of applying rotary embedding to query and key, and is thus much more efficient than the PyTorch implementation.
Tested models:
Tested GPUs: