Add custom kernel for RMS normalization by WoosukKwon · Pull Request #16 · vllm-project/vllm

WoosukKwon · 2023-03-31T01:24:40Z

This PR adds a custom CUDA kernel for RMS normalization, which is used in LLaMA models. The kernel removes the inefficient data movement in the current PyTorch implementation.

Performance (fp16):

* num_tokens=7, hidden_size=1024
Kernel: 9 us
PyTorch: 93 us

* num_tokens=128, hidden_size=1024
Kernel: 5 us
PyTorch: 84 us

* num_tokens=2048, hidden_size=5120
Kernel: 60 us
PyTorch: 353 us

Tested models:

LLaMA-7B
LLaMA-13B

Tested GPUs:

A100

zhuohan123

LGTM

Generalizing KV scales JSON to updated schema

Extra optimizations for PA

chore: add OWNERS file to ibm_main

Fix ambiguous fma call

SUMMARY: - Fix bug whereby 2:4 is not being invoked - Eschew SparseTensor based implementation TESTING: - examples/offline_inference_semi_structured_sparse.py --------- Co-authored-by: Lucas Wilkinson <wilkinson.lucas@gmail.com>

…IO. (vllm-project#16…" This reverts commit 68d4c33.

Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com>

* [Update] LMcache connector v1 implementation Signed-off-by: ApostaC <yihua98@uchicago.edu> * [Add] examples for disaggregated prefill Signed-off-by: ApostaC <yihua98@uchicago.edu> * [add] extra information about evns Signed-off-by: ApostaC <yihua98@uchicago.edu> * Initial stubs for P/D scheduling changes Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * Updates Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * Rs branch (#3) * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * Rs branch (#5) Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * Remove Unneeded Arguments (#7) * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * stash Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * cleanup Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> --------- Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * Improve disagg-example.sh (#8) - fix spelling - CUDA_VISIBLE_DEVICES should be set externally Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * added connector Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * update Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * remove Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * seems to load properly Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * Revert "updated" This reverts commit 97316d9. * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * stash Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * added Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * diffs for local dev on macos Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * update Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updaed Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * Checkpoint. Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * Cleanup Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * WIP Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated on scheduler side Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * Hacking away Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * cleanup Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * ensure request removed from running list Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * Runs E2E. Garbage output. Crashes on 2nd request Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * update Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * rename files Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * update Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * Second request no longer crashes Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * Remove gpu_model_runner hacks Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * Clean up Justfile Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * [Bugfix] Stale finished requests in EMPTY_MODEL_RUNNER_OUTPUT Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * update Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * justfile edits Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * Update Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * Fixes - lm_eval gsm8k has correctness Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * "just delete the assert" Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * fixup precommit issues Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * Fixes Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * updated (#12) Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * Add Accuracy Test (#13) * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> --------- Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * Preemption Bugfixes (#15) * stash fixed double free issue Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * fixed issue Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updatrd Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updatrd Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updatrd Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updatrd Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updatrd Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updatrd Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> --------- Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated (#16) Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * Fix Bad Merge | Fix Memory Leak in Upstream (#18) * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * fix merge Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> --------- Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * clean up justfile, examples Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * more cleanup Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * more cleanup Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * more cleanup Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * more cleanup Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * More cleanup Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * more cleanup Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * more cleanup, precommit fixes Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * More cleanup Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * run_accuracy_test.sh UX Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * squash warnings Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * pre-commit Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * update Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * Add get_finished to base kv connector Signed-off-by: mgoin <mgoin64@gmail.com> * revert test.txt Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * cleanup Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * Cleanup Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * review comments Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> --------- Signed-off-by: ApostaC <yihua98@uchicago.edu> Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> Signed-off-by: mgoin <mgoin64@gmail.com> Co-authored-by: ApostaC <yihua98@uchicago.edu> Co-authored-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com> Co-authored-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> Co-authored-by: Robert Shaw <rshaw@neuralmagic.com> Co-authored-by: mgoin <mgoin64@gmail.com> Co-authored-by: mgoin <mgoin64@gmail.com>

* [Update] LMcache connector v1 implementation Signed-off-by: ApostaC <yihua98@uchicago.edu> * [Add] examples for disaggregated prefill Signed-off-by: ApostaC <yihua98@uchicago.edu> * [add] extra information about evns Signed-off-by: ApostaC <yihua98@uchicago.edu> * Initial stubs for P/D scheduling changes Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * Updates Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * Rs branch (#3) * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * Rs branch (#5) Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * Remove Unneeded Arguments (#7) * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * stash Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * cleanup Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> --------- Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * Improve disagg-example.sh (#8) - fix spelling - CUDA_VISIBLE_DEVICES should be set externally Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * added connector Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * update Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * remove Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * seems to load properly Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * Revert "updated" This reverts commit 97316d9. * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * stash Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * added Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * diffs for local dev on macos Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * update Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updaed Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * Checkpoint. Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * Cleanup Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * WIP Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated on scheduler side Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * Hacking away Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * cleanup Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * ensure request removed from running list Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * Runs E2E. Garbage output. Crashes on 2nd request Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * update Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * rename files Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * update Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * Second request no longer crashes Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * Remove gpu_model_runner hacks Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * Clean up Justfile Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * [Bugfix] Stale finished requests in EMPTY_MODEL_RUNNER_OUTPUT Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * update Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * justfile edits Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * Update Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * Fixes - lm_eval gsm8k has correctness Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * "just delete the assert" Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * fixup precommit issues Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * Fixes Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * updated (#12) Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * Add Accuracy Test (#13) * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> --------- Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * Preemption Bugfixes (#15) * stash fixed double free issue Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * fixed issue Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updatrd Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updatrd Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updatrd Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updatrd Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updatrd Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updatrd Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> --------- Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated (#16) Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * Fix Bad Merge | Fix Memory Leak in Upstream (#18) * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * fix merge Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> --------- Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * cleanup code Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * cleanup code Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * stash Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updatted Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * revert Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * more spurious changes Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * Update vllm/distributed/kv_transfer/kv_connector/v1/nixl_connector.py Co-authored-by: Tyler Michael Smith <tysmith@redhat.com> * Update vllm/distributed/kv_transfer/kv_connector/v1/nixl_connector.py Co-authored-by: Tyler Michael Smith <tysmith@redhat.com> --------- Signed-off-by: ApostaC <yihua98@uchicago.edu> Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> Co-authored-by: ApostaC <yihua98@uchicago.edu> Co-authored-by: Tyler Michael Smith <tyler@neuralmagic.com> Co-authored-by: Tyler Michael Smith <tysmith@redhat.com> Co-authored-by: Robert Shaw <rshaw@neuralmagic.com>

* [Update] LMcache connector v1 implementation Signed-off-by: ApostaC <yihua98@uchicago.edu> * [Add] examples for disaggregated prefill Signed-off-by: ApostaC <yihua98@uchicago.edu> * [add] extra information about evns Signed-off-by: ApostaC <yihua98@uchicago.edu> * Initial stubs for P/D scheduling changes Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * Updates Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * Rs branch (#3) * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * Rs branch (#5) Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * Remove Unneeded Arguments (#7) * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * stash Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * cleanup Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> --------- Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * Improve disagg-example.sh (#8) - fix spelling - CUDA_VISIBLE_DEVICES should be set externally Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * added connector Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * update Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * remove Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * seems to load properly Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * Revert "updated" This reverts commit 97316d9. * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * stash Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * added Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * diffs for local dev on macos Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * update Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updaed Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * Checkpoint. Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * Cleanup Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * WIP Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated on scheduler side Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * Hacking away Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * cleanup Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * ensure request removed from running list Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * Runs E2E. Garbage output. Crashes on 2nd request Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * update Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * rename files Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * updated Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * update Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> * Second request no longer crashes Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * Remove gpu_model_runner hacks Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * Clean up Justfile Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * [Bugfix] Stale finished requests in EMPTY_MODEL_RUNNER_OUTPUT Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * update Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * justfile edits Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * Update Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * Fixes - lm_eval gsm8k has correctness Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * "just delete the assert" Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * fixup precommit issues Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * Fixes Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * updated (#12) Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * Add Accuracy Test (#13) * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> --------- Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * Preemption Bugfixes (#15) * stash fixed double free issue Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * fixed issue Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updatrd Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updatrd Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updatrd Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updatrd Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updatrd Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updatrd Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> --------- Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated (#16) Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * Fix Bad Merge | Fix Memory Leak in Upstream (#18) * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * fix merge Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> --------- Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * cleanup code Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * cleanup code Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * stash Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updatted Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * revert Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * more spurious changes Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * updated Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> * Support MLA in NIXL connector Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * WIP adding tests Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * wip Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> * Fixes Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> --------- Signed-off-by: ApostaC <yihua98@uchicago.edu> Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com> Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> Signed-off-by: Robert Shaw <rshaw@neuralmagic.com> Co-authored-by: ApostaC <yihua98@uchicago.edu> Co-authored-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com> Co-authored-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com> Co-authored-by: Robert Shaw <rshaw@neuralmagic.com>

…/mid prompts), de-spilled QSA kernel (opt-in, no in-server gain), int8 AR own-shard in place; trace_agg families VLLM_RDNA_GEMM_MBUCKET=128: odd-M prefill steps (block-boundary stops) no longer fall back to rocBLAS MT32x32x8; TunableOp rows for M=384..2048. VLLM_RDNA_QSA_DESPILL=1: 1.17x harness, ~0 in-server. CHANGES vllm-project#16, including a stability note: every bus drop since 09-25 had VLLM_RDNA_AR_Q8=1. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

RELEASE-ONLY. The CPU test shards do not pull the image published by image_build_cpu.sh -- run-cpu-test.sh builds its own image with a plain `docker build`, using the agent's local BuildKit cache. So the `--no-cache` added to image_build_cpu.sh does not cover them. torch is installed in the `base-common` stage of docker/Dockerfile.cpu, which is upstream of the vLLM source COPY, so its layer key depends only on the base image and `requirements/cpu.txt`. Neither changes when the 2.14.0 RC is respun under the same version string, so the cached layer is restored every time. In build 84719 the `CPU-Multi-Modal Model Tests 1` shard reported steps vllm-project#16, vllm-project#6, wheel and could not have exercised the ideep/oneDNN revert. Add `--no-cache` to the shard build, and set `UV_NO_CACHE=1` in the Dockerfile base stage: `--no-cache` re-executes the layer but does not clear the `RUN --mount=type=cache,target=/root/.cache/uv` mount, so uv would still serve the stale same-named wheel from its own cache. Revert both once 2.14.0 is final and published to PyPI.

A corruption rejection that evicts a chunk another request's lookup already pinned crashed the worker. CPUOffloadingManager._prepare_load degraded the vanished key to a miss and returned a spec shorter than the GPU block list the offloading scheduler paired in update_state_after_alloc, so transfer planning asserted at gpu_worker.py:653 inside start_kv_transfers. Had the worker survived, the job's completion would have asserted again in complete_load, whose key set still held the vanished key. The defect dates from the lookup-pin degrade path introduced in vllm-project#16. prepare_load now pre-scans for a pinned key that is no longer in the cache and raises OffloadKeyVanishedError before consuming any key, so earlier keys keep their pins and ref_cnt. The connector scheduler catches it, builds no load job and carries the request id to the workers as a failed load in the connector metadata. Each worker rank reports that id in both finished_recving and failed_recving through get_transfer_results: KVOutputAggregator only surfaces a failure that every rank also reports as finished, and the core scheduler's _handle_failed_recving acts on it because the request is already in WAITING_FOR_REMOTE_KVS. No TransferJobStatus exists for the request, so complete_load never sees the vanished key. The block-level channel, get_block_ids_with_load_errors, is not usable here: the core scheduler raises for block-level load failures on models with more than one KV cache group, and the hybrid models served by this connector have several. With the default kv_load_failure_policy="fail" the request finishes with an error and the engine keeps serving. With "recompute" the same report zeroes the request's computed tokens and reschedules it silently. The retry's lookup answers MISS for the evicted chunk. The pins now survive to request finish instead of being dropped by the degrade, so on_request_finished settles each confirmer's share and its stale-pin skip tolerates the vanished key. The share that previously stranded in _lookup_pinned when two confirmers degraded is gone. It also adds the vanished-pin regression test that drives the real manager, scheduler and worker planning path. The test was first written red against the old short-spec crash and is landed here against the failed-receive contract. Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> Signed-off-by: Adrian Zankich <adzankich@gmail.com>

…llers [CI][Lwd] Wire the shared LWD CI workflows into the repository

…roject#16) Restore LaunchSpec and JitKernel infrastructure lost during the port. Cover the warmup and compile paths, factory returns, arguments, and stride ownership. Preserve the reviewed source tree from PR vllm-project#16; the squash only reorganizes Git history. Existing acceptance evidence and remaining limitations are recorded in overview PR vllm-project#11. No new model tests were run for this integration. Original-PR-Head: 9f6afed7f84ca0b615029824423e94b870aed8dc Assisted-by: Codex

Skip empty query windows when gathering multimodal embeddings. PCP keeps a zero-length placeholder row on ranks without tokens; when its cursor falls inside a media feature, the old gather path asserted on the empty intersection. Allow multimodal PCP for GLM-5.3-Flash without M-RoPE. Keep other model configurations gated because M-RoPE still uses global request cursors instead of PCP segment starts. Validation: - Encoder runner and PCP manager suites: 121 passed, 5 skipped. - Multimodal configuration scope checks rerun: 3 passed. - Real image requests on 8x S5000, eager, FP8 KV, 8K max model length: TP8/PCP1 and TP1/PCP8 with KDA TP8 each completed all 7 requests. Six cases were token-identical; the scene description differed only by "weathered", with tied candidates at the first baseline divergence. - A 2055-token real image prompt with a 7-token final prefill chunk starting inside the image completed on PCP8 and returned STOP. Co-authored-by: Codex <noreply@openai.com>

…roject#16) Restore LaunchSpec and JitKernel infrastructure lost during the port. Cover the warmup and compile paths, factory returns, arguments, and stride ownership. Preserve the reviewed source tree from PR vllm-project#16; the squash only reorganizes Git history. Existing acceptance evidence and remaining limitations are recorded in overview PR vllm-project#11. No new model tests were run for this integration. Original-PR-Head: 9f6afed7f84ca0b615029824423e94b870aed8dc Assisted-by: Codex

Skip empty query windows when gathering multimodal embeddings. PCP keeps a zero-length placeholder row on ranks without tokens; when its cursor falls inside a media feature, the old gather path asserted on the empty intersection. Allow multimodal PCP for GLM-5.3-Flash without M-RoPE. Keep other model configurations gated because M-RoPE still uses global request cursors instead of PCP segment starts. Validation: - Encoder runner and PCP manager suites: 121 passed, 5 skipped. - Multimodal configuration scope checks rerun: 3 passed. - Real image requests on 8x S5000, eager, FP8 KV, 8K max model length: TP8/PCP1 and TP1/PCP8 with KDA TP8 each completed all 7 requests. Six cases were token-identical; the scene description differed only by "weathered", with tied candidates at the first baseline divergence. - A 2055-token real image prompt with a 7-token final prefill chunk starting inside the image completed on PCP8 and returned STOP. Co-authored-by: Codex <noreply@openai.com>

…roject#16) Restore LaunchSpec and JitKernel infrastructure lost during the port. Cover the warmup and compile paths, factory returns, arguments, and stride ownership. Preserve the reviewed source tree from PR vllm-project#16; the squash only reorganizes Git history. Existing acceptance evidence and remaining limitations are recorded in overview PR vllm-project#11. No new model tests were run for this integration. Original-PR-Head: 9f6afed7f84ca0b615029824423e94b870aed8dc Assisted-by: Codex

Skip empty query windows when gathering multimodal embeddings. PCP keeps a zero-length placeholder row on ranks without tokens; when its cursor falls inside a media feature, the old gather path asserted on the empty intersection. Allow multimodal PCP for GLM-5.3-Flash without M-RoPE. Keep other model configurations gated because M-RoPE still uses global request cursors instead of PCP segment starts. Validation: - Encoder runner and PCP manager suites: 121 passed, 5 skipped. - Multimodal configuration scope checks rerun: 3 passed. - Real image requests on 8x S5000, eager, FP8 KV, 8K max model length: TP8/PCP1 and TP1/PCP8 with KDA TP8 each completed all 7 requests. Six cases were token-identical; the scene description differed only by "weathered", with tied candidates at the first baseline divergence. - A 2055-token real image prompt with a 7-token final prefill chunk starting inside the image completed on PCP8 and returned STOP. Co-authored-by: Codex <noreply@openai.com>

The debug-only DSA indexer top-k count checks were unused and the prefill check fired spuriously under DCP (it ran after the global merge but compared against per-rank lengths). Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

WoosukKwon added 4 commits March 31, 2023 01:15

Add reduction_utils.h

3d17680

Add RMS norm kernel

2a3f6eb

Add tests for RMS norm kernel

a6b924e

Add RMSNorm module

9921699

WoosukKwon requested a review from zhuohan123 March 31, 2023 01:25

zhuohan123 approved these changes Mar 31, 2023

View reviewed changes

zhuohan123 merged commit 09e9245 into main Mar 31, 2023

WoosukKwon deleted the rms-norm branch March 31, 2023 16:51

shanshanpt mentioned this pull request Nov 17, 2023

Run long conetxt error : CUDA error: an illegal memory access was encountered #1700

Closed

junior-zsy mentioned this pull request Nov 20, 2023

Error with 32k Long Text in chatglm2-6b-32k Model #1725

Closed

hongxiayang pushed a commit to hongxiayang/vllm that referenced this pull request Feb 13, 2024

Add custom kernel for RMS normalization (vllm-project#16)

3d52973

AdrianAbeyta added a commit to AdrianAbeyta/vllm that referenced this pull request Mar 8, 2024

Merge pull request vllm-project#16 from ROCm/fp8_ingest_stage1_model

7b72159

Generalizing KV scales JSON to updated schema

luo-cheng2021 pushed a commit to luo-cheng2021/vllm that referenced this pull request Apr 1, 2024

Merge pull request vllm-project#16 from ilya-lavrenov/pa-optimizations

2ce2ff7

Extra optimizations for PA

z103cb added a commit to dtrifiro/vllm that referenced this pull request May 8, 2024

Merge pull request vllm-project#16 from z103cb/update-ibm-main-owners

8a6c9c9

chore: add OWNERS file to ibm_main

fxmarty pushed a commit to fxmarty/vllm-public that referenced this pull request May 31, 2024

Merge pull request vllm-project#16 from cjatin/bfloat16_fix

24584bc

Fix ambiguous fma call

yuhuixu1993 mentioned this pull request Jun 2, 2024

[Bug]: loading squeezellm model #5190

Closed

alixiaodi mentioned this pull request Aug 2, 2024

[Bug]: #7072

Closed

SpaceHunterInf mentioned this pull request Sep 30, 2024

[Bug]: Bus error (core dumped) #8974

Closed

1 task done

ifoyil mentioned this pull request Mar 3, 2025

[Installation]: Dockerfile.cpu installation problem vLLM #14033

Closed

1 task done

chaunceyjiang added a commit to chaunceyjiang/vllm that referenced this pull request Apr 23, 2025

Revert "[Misc] Add S3 environment variables for better support of Min…

710d556

…IO. (vllm-project#16…" This reverts commit 68d4c33.

robertgshaw2-redhat referenced this pull request in robertgshaw2-redhat/vllm May 1, 2025

updated (#16)

a45a694

Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com>

hao-cold mentioned this pull request May 13, 2025

[Bug]: CUDA error: an illegal instruction was encountered #18045

Closed

1 task done

markmc mentioned this pull request May 21, 2025

[Bug][Failing Test]: Distributed Comm Ops - distributed/test_shm_broadcast.py #18492

Closed

1 task done

pdtgct mentioned this pull request Oct 18, 2026

[Proposal] Move cache-aware streaming ASR state and the Nemotron streaming ASR model from vLLM-Omni to vLLM core pdtgct/vllm#1

Open

ISEEKYAN mentioned this pull request Oct 21, 2026

[DSv4.1][BI] Bitwise batch invariance for DeepSeek-V4.1-Flash with CUDA graphs #62365

Open

tjtanaa mentioned this pull request Oct 26, 2026

[RFC]: Attention Backend Test Suites #63640

Open

1 task done

Chloe-orz pushed a commit to Chloe-orz/vllm-maintomain that referenced this pull request Oct 26, 2026

Merge pull request vllm-project#16 from vllm-data-comm/feat/lwd-ci-ca…

fe3b820

…llers [CI][Lwd] Wire the shared LWD CI workflows into the repository

Navigation Menu

Repository navigation

Add custom kernel for RMS normalization - #16

Add custom kernel for RMS normalization#16
zhuohan123 merged 4 commits into
mainfrom
rms-norm

WoosukKwon commented Mar 31, 2023

Uh oh!

zhuohan123 left a comment

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

WoosukKwon commented Mar 31, 2023

Uh oh!

zhuohan123 left a comment

Choose a reason for hiding this comment

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants