Repository navigation
Merge QKV into one linear layer - #15
Merged
Merged
Conversation
WoosukKwon
reviewed
Mar 30, 2023
WoosukKwon
left a comment
WoosukKwon
left a comment
Collaborator
There was a problem hiding this comment.
Thanks for your effort. Please check my comments.
Merged
Collaborator
WoosukKwon
self-requested a review
April 2, 2023 07:23
WoosukKwon
approved these changes
Apr 2, 2023
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
slyalin
pushed a commit
to slyalin/vllm
that referenced
this pull request
Mar 26, 2024
…no-model-executor-opt [CPU] Avoid copy result and force allocation
z103cb
referenced
this pull request
in z103cb/opendatahub_vllm
Apr 22, 2024
This PR updates our grpc_server to add TGIS-style logs similar to https://github.com/IBM/text-generation-inference/blob/main/router/src/grpc_server.rs#L504-L512 This also disables the vllm per-request logging so that we don't double-log each request The timing info collected here is pretty rough, it doesn't plumb into the LLMEngine, it just times the generators to get the total time spent in the engine. We could do better, but this is a start. Example logs: ``` INFO 04-09 21:51:01 logs.py:43] generate_stream{input=[b'This is the story of Obama ridin...'] prefix_id= input_chars=[70] params=sampling { } stopping { max_new_tokens: 200 min_new_tokens: 16 } response { } decoding { } tokenization_time=0.45ms queue_and_inference_time=1096.67ms time_per_token=5.48ms total_time=1097.12ms input_toks=16}: Streaming response generated 200 tokens before NOT_FINISHED, output 848 chars: b' California. The story is told i...' INFO 04-09 21:51:08 logs.py:43] generate{input=[b'Lorem ipsum dolor sit amet, cons...', b'foooood man where is it'] prefix_id= input_chars=[469] params=sampling { } stopping { max_new_tokens: 20 min_new_tokens: 16 } response { } decoding { } tokenization_time=2.03ms queue_and_inference_time=122.23ms time_per_token=6.11ms total_time=124.26ms input_toks=124}: Sub-request 0 from batch of 2 generated 20 tokens before MAX_TOKENS, output 25 chars: b'?\\n\\n<!--\\n<!--\\n<!--\\n<!--\\n<!' INFO 04-09 21:51:08 logs.py:43] generate{input=[b'Lorem ipsum dolor sit amet, cons...', b'foooood man where is it'] prefix_id= input_chars=[469] params=sampling { } stopping { max_new_tokens: 20 min_new_tokens: 16 } response { } decoding { } tokenization_time=2.07ms queue_and_inference_time=122.22ms time_per_token=6.11ms total_time=124.29ms input_toks=7}: Sub-request 1 from batch of 2 generated 20 tokens before MAX_TOKENS, output 70 chars: b"?\\nI don't know.\\nI don't know.\\nI ..." ``` --------- Signed-off-by: Joe Runde <Joseph.Runde@ibm.com> Signed-off-by: Joe Runde <joe@joerun.de>
fxmarty
pushed a commit
to fxmarty/vllm-public
that referenced
this pull request
May 31, 2024
Correctly calculating the same value for the required cache blocks num for all torchrun processes
ykim362
pushed a commit
to ykim362/vllm
that referenced
this pull request
Jun 17, 2024
…-wenxh/fp8-on-a100-v5-pr Revert "0612 kernel of FP8 on A100"
Closed
1 task done
robertgshaw2-redhat
referenced
this pull request
in robertgshaw2-redhat/vllm
May 1, 2025
* 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>
robertgshaw2-redhat
referenced
this pull request
in robertgshaw2-redhat/vllm
May 3, 2025
* [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>
robertgshaw2-redhat
referenced
this pull request
in robertgshaw2-redhat/vllm
May 4, 2025
* [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>
robertgshaw2-redhat
referenced
this pull request
in robertgshaw2-redhat/vllm
May 6, 2025
* [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>
itej89
pushed a commit
to itej89/vllm
that referenced
this pull request
Jul 20, 2026
heliubj18
added a commit
to heliubj18/vllm
that referenced
this pull request
Aug 7, 2026
Two things learned from build vllm-project#15, which took 8 passed / 7 failed and was cancelled by hand with 27 jobs still queued. Fail fast. GPU steps here all contend for one box under gpu_slots: 1, so a step queued behind a failure holds the only slot to produce a verdict for a run that is already red - and the tail is long, with Sequence Parallel alone at 41 minutes. Emit cancel_on_build_failing on every generated step, behind a fail_fast config key so it can be turned off when a full inventory of breakage in one pass is worth more than the time. Drop cupy-cuda12x from the venv. ray[cgraph] depends on it, the image is CUDA 13 and ships cupy-cuda13x, and PYTHONPATH puts the venv's copy first - so cupy loaded a CUDA 12 build against a CUDA 13 runtime, warning about it and continuing. Self-inflicted by adding ray, hence fixed in the setup script rather than by hand on the host. The env comment records something that is not a fix: removing the pipeline-wide VLLM_WORKER_MULTIPROC_METHOD=spawn does not repair Model Runner V2 Distributed, which fails with "Cannot re-initialize CUDA in forked subprocess" with or without it. Leaving it unset still matches upstream, which exports it per-step, but the real cause there is a WorkerProc background exception and is still open. Co-authored-by: Claude Opus 5
heliubj18
added a commit
to heliubj18/vllm
that referenced
this pull request
Aug 7, 2026
…e device Upstream's CPU markings mean "performs no GPU computation", not "runs without a visible device". Importing vllm loads _C_stable_libtorch.abi3.so, which needs libcuda.so.1, so in a GPU-less container the import fails and the tests degrade into "Failed to infer device type" and pydantic ValidationErrors on ModelConfig rather than anything naming the real problem. Build vllm-project#15's V1 Others (CPU) came back 194 failed / 208 passed this way, and Basic Models Test (Other CPU) failed at collection with ModuleNotFoundError: No module named 'vllm._C'. Upstream is unaffected because its CPU steps run on a CPU-built image; we have one CUDA image for everything. A controlled run over v1/core/test_worker_slot_overflow.py shows both the devices and the shared read-only HF cache mount are required - neither alone is enough: no GPU, no cache 3 failed no GPU, cache 3 failed GPU, no cache 3 failed GPU + cache 3 passed These steps keep their cpu_slots concurrency group rather than taking the single GPU slot: VRAM use was measured at 1 MiB while running, so they only need the device present, not idle. Co-authored-by: Claude Opus 5
heliubj18
added a commit
to heliubj18/vllm
that referenced
this pull request
Aug 7, 2026
Build vllm-project#15 established that Sequence Parallel Correctness Tests passes, at a cost of 41 minutes. Re-proving that on every iteration of the remaining fixes is the bulk of the wall clock, so use the step_allowlist debugging hook to run only what is being verified. fail_fast goes off for the same reason: six independent fixes are in flight and cancelling the rest on the first failure would hide whether the others landed. engine-1-gpu stays out of the list because it needs meta-llama/Llama-3.2-1B as TOKENIZER_NAME and that repo answers 403 until access is approved on the Hub - running it now only re-proves a known blocker. Both settings revert together once these six are green. Co-authored-by: Claude Opus 5
heliubj18
added a commit
to heliubj18/vllm
that referenced
this pull request
Aug 8, 2026
Every step the allowlist was ever narrowed to is now accounted for, so the debugging aid has served its purpose: kernels-attention-test passed 119m25s in vllm-project#17 kernels-quantization-test passed 110m25s in vllm-project#21 kernels-moe-test passed 223m09s in vllm-project#21 sequence-parallel passed 41m32s in vllm-project#15 v1-others-cpu 402 cases / 1 failure in vllm-project#23 engine-1-gpu denylisted, upstream fork bug model-runner-v2-distributed denylisted, same bug model-runner-v2-distributed-2-gpus joins the denylist. It is the same flashinfer-poisons-fork bug as engine-1-gpu, reached through test_basic_correctness.py:23 instead: importing tests/utils runs the chain, then tests/utils.py:1745 forks, and the MultiProcExecutor worker in the child dies at gpu_worker.py:369. The log carries the fork DeprecationWarning at that exact line. Keeping it was tempting because only the first of its three commands is affected - the two DP commands do not import tests/utils and both pass by hand (test_async_llm_dp 14 selected of 18, test_eagle_dp 1 passed in 145s). But they have never run in CI: the first command always crashes first and a Buildkite step is atomic. So keeping the step buys no coverage, only a red build 43 seconds in, which under fail_fast would cancel every GPU step behind it. Recovering those two needs command-level filtering, which the generator does not have. v1-others-cpu is left in: its 30 missing-repo failures are fixed now that Qwen3-0.6B, Qwen1.5-7B and Qwen2.5-VL-3B-Instruct are staged as config-only (12M each - they are constructed through ModelConfig(...), not loaded). Its one remaining failure wants world size 2 while holding one card; upstream runs that step on a CPU image where ParallelConfig does not check GPU count. That is an image-type difference, not a card allocation bug, and it is not worth a per-key exception. Co-authored-by: Claude Opus 5
djdisodo
pushed a commit
to djdisodo/vllm
that referenced
this pull request
Aug 13, 2026
…-project#15) Three independent fixes for the flat aggregate-throughput report on Qwen3-4B. Quant numerics are unchanged. 1. Vectorized the tile flush. _batched_flush gathered and wrote tiles in a per-(layer,block,head) Python loop, exploding into ~1e5 tiny GPU ops on a synchronized burst (prefill completion, lockstep decode boundary) and dominating the metadata build at high concurrency. Replaced with one index_select gather + one index_copy write per (layer, block-chunk), reusing the exact same Sinkhorn and RTN/pack calls and the same packed byte layout, so output is bit-for-bit identical (validated by greedy token match vs the legacy path). Legacy path kept behind KVARN_FAST_FLUSH=0 for A/B. 2. Weight-aware tail-pool budget. The pool was sized as a fixed fraction of TOTAL device memory, capping concurrency while the KV cache sat near-empty (a 4B on a 24 GB card was capped to ~22 seqs with most of the usable memory unused). Size it instead from the post-weight usable envelope, share*(util*total - weights), reading weight bytes off the checkpoint files on disk (no device context, in the early platform hook), with a safe fallback to the legacy fraction when weights cannot be read. 3. Flush-kernel warmup. The Sinkhorn and int4-store kernels only run at a tile-boundary flush, which never happens during the profiling/dummy run, so they JIT-compiled mid-serving (a latency spike and a jit_monitor warning). Compile them once at pool init on throwaway tiles.
djdisodo
pushed a commit
to djdisodo/vllm
that referenced
this pull request
Aug 13, 2026
…ost) The variance-normalization converges by ~4 iterations and the kernel tracks the best-so-far scales, so 8 iterations reconstruct the same KV as 16. Validated token-identical on Qwen3-4B / AIME25 in fp16: iters=8 and iters=16 give the same per-seed scores (avg@2 45.0%). Halving the iteration count roughly halves the Sinkhorn flush kernel, which is the dominant KVarN-specific decode-GPU cost at burst (issue vllm-project#15), lifting aggregate throughput at N=128 from ~3.5k to ~4.0k tok/s on a 4B (clearing the TurboQuant baseline) with no accuracy change. Env-overridable via KVARN_SINKHORN_ITERS; raise it for large reasoning models if a convergence check shows benefit. Validated on one model (Qwen3-4B); other families inherit the lower default and can override if needed.
djdisodo
pushed a commit
to djdisodo/vllm
that referenced
this pull request
Aug 14, 2026
…-project#15) Three independent fixes for the flat aggregate-throughput report on Qwen3-4B. Quant numerics are unchanged. 1. Vectorized the tile flush. _batched_flush gathered and wrote tiles in a per-(layer,block,head) Python loop, exploding into ~1e5 tiny GPU ops on a synchronized burst (prefill completion, lockstep decode boundary) and dominating the metadata build at high concurrency. Replaced with one index_select gather + one index_copy write per (layer, block-chunk), reusing the exact same Sinkhorn and RTN/pack calls and the same packed byte layout, so output is bit-for-bit identical (validated by greedy token match vs the legacy path). Legacy path kept behind KVARN_FAST_FLUSH=0 for A/B. 2. Weight-aware tail-pool budget. The pool was sized as a fixed fraction of TOTAL device memory, capping concurrency while the KV cache sat near-empty (a 4B on a 24 GB card was capped to ~22 seqs with most of the usable memory unused). Size it instead from the post-weight usable envelope, share*(util*total - weights), reading weight bytes off the checkpoint files on disk (no device context, in the early platform hook), with a safe fallback to the legacy fraction when weights cannot be read. 3. Flush-kernel warmup. The Sinkhorn and int4-store kernels only run at a tile-boundary flush, which never happens during the profiling/dummy run, so they JIT-compiled mid-serving (a latency spike and a jit_monitor warning). Compile them once at pool init on throwaway tiles.
djdisodo
pushed a commit
to djdisodo/vllm
that referenced
this pull request
Aug 14, 2026
…ost) The variance-normalization converges by ~4 iterations and the kernel tracks the best-so-far scales, so 8 iterations reconstruct the same KV as 16. Validated token-identical on Qwen3-4B / AIME25 in fp16: iters=8 and iters=16 give the same per-seed scores (avg@2 45.0%). Halving the iteration count roughly halves the Sinkhorn flush kernel, which is the dominant KVarN-specific decode-GPU cost at burst (issue vllm-project#15), lifting aggregate throughput at N=128 from ~3.5k to ~4.0k tok/s on a 4B (clearing the TurboQuant baseline) with no accuracy change. Env-overridable via KVARN_SINKHORN_ITERS; raise it for large reasoning models if a convergence check shows benefit. Validated on one model (Qwen3-4B); other families inherit the lower default and can override if needed.
BugenZhao
added a commit
to biswapanda/vllm
that referenced
this pull request
Sep 24, 2026
Use the upstream repository at f0985ef65967615db2c79279aa07818499301bfd after llm-multimodal vllm-project#15 merged. Co-authored-by: Codex <noreply@openai.com> Signed-off-by: Bugen Zhao <i@bugenzhao.com>
leapdragon
added a commit
to leapdragon/vllm-rdna2-qwen
that referenced
this pull request
Sep 26, 2026
…no GPU peer-to-peer Payload through one shared pinned host buffer with LL-style tagged 8-byte words (correct regardless of write ordering); one poller per block per peer; fixed-order fp32 reduction, graph-replay safe, T44b abort handling unchanged. Op test on 4x V620: 32.1 us eager / 33.3 us in-graph per 20 KB (p2p kernel 31-33, RCCL ~156), bit-identical across ranks. Motivation: the p2p kernel's small BAR writes coincide with cards dropping off the bus here and on users' boards; VLLM_RDNA_AR=0 costs 26 % decode. Default stays p2p. Also: cmake/hipify.py resolves symlinks consistently (symlinked checkout left cuda_compat.h un-hipified); ar_ops_test.py can test a built library without installing it. CHANGES vllm-project#15, TROUBLESHOOTING 5d. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
1 task done
Chloe-orz
pushed a commit
to Chloe-orz/vllm-maintomain
that referenced
this pull request
Oct 26, 2026
…ones Use the PR template from AGENTS.md section 5 verbatim, so new PRs start from the team checklist instead of the upstream vLLM one. Drop the eight upstream issue forms in favor of three plain Markdown templates: Bug, RFC and Idea. The RFC template keeps the [RFC] title prefix because the ci-infra rfc-link check identifies RFC issues by it. Remove the contact link to the upstream vLLM forum. Closes vllm-project#15 Signed-off-by: tu10ng <2059734099@qq.com>
This was referenced Oct 30, 2026
qianlihuang
added a commit
to qianlihuang/vllm
that referenced
this pull request
Oct 30, 2026
Kimi Linear/K3 prefill ran three separate causal convolutions over the Q, K and V slices. The conv weights and conv state are already stored packed, so run a single convolution over the packed q|k|v input and split the output into views, which FlashKDA vllm-project#15 consumes directly. The raw projection is kept for the checkpoint exporter. Signed-off-by: qianlihuang <yiliu.dong@qq.com> Co-Authored-By: Claude Code <noreply@anthropic.com>
qianlihuang
added a commit
to qianlihuang/vllm
that referenced
this pull request
Oct 31, 2026
Kimi Linear/K3 prefill ran three separate causal convolutions over the Q, K and V slices. The conv weights and conv state are already stored packed, so run a single convolution over the packed q|k|v input and split the output into views, which FlashKDA vllm-project#15 consumes directly. The raw projection is kept for the checkpoint exporter. Signed-off-by: qianlihuang <yiliu.dong@qq.com> Co-Authored-By: Claude Code <noreply@anthropic.com>
KexyBiscuit
pushed a commit
to tai-core/vllm-tracking
that referenced
this pull request
Nov 4, 2026
…-project#15) Route ordinary decode through recurrent KDA and prefill through chunked KDA in non-speculative mixed batches. Preserve state and output ownership and add precision migration contracts. Preserve the reviewed source tree from PR vllm-project#15; 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: 3e4b131acd5c1d3357513404d29ef635b0a1edb2 Assisted-by: Codex
KexyBiscuit
pushed a commit
to tai-core/vllm-tracking
that referenced
this pull request
Nov 4, 2026
…-project#15) Route ordinary decode through recurrent KDA and prefill through chunked KDA in non-speculative mixed batches. Preserve state and output ownership and add precision migration contracts. Preserve the reviewed source tree from PR vllm-project#15; 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: 3e4b131acd5c1d3357513404d29ef635b0a1edb2 Assisted-by: Codex
KexyBiscuit
pushed a commit
to tai-core/vllm-tracking
that referenced
this pull request
Nov 4, 2026
…-project#15) Route ordinary decode through recurrent KDA and prefill through chunked KDA in non-speculative mixed batches. Preserve state and output ownership and add precision migration contracts. Preserve the reviewed source tree from PR vllm-project#15; 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: 3e4b131acd5c1d3357513404d29ef635b0a1edb2 Assisted-by: Codex
triangleXIV
pushed a commit
to tokmo-infra/vllm
that referenced
this pull request
Nov 9, 2026
vllm-project#15) * [Bugfix][Kimi K3.1] Review fixes: registry entry, PP wiring check, Mooncake prefix-caching guard, cleanups - Add KimiK31ForConditionalGeneration to tests/models/registry.py (fixes test_hf_registry_coverage) and to supported_models.md. - DSA wiring check: count only this PP stage's layers; other stages are PPMissingLayer and have no self_attn. - Mooncake: reject a consumer with prefix caching whenever the logical block sizes differ, not only for hybrid models; otherwise the producer writes the prompt's leading blocks into the consumer's post-hit blocks. - Remove unused state and duplicate checks: FlashMLASparseMetadataBuilder .use_merged_cache, KimiDecoderLayer._self_attn_writes_output, the unused attn_backend parameter of the K3 MLA layer, and the CUDA selector's sparse filter (validate_configuration already rejects it). - Merge the two KimiK3DSAAttention construction branches, make split_merged_dsa_cache_views public, and trim comments. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * [Doc] Kimi K3.1: use the moonshotai/Kimi-K3.1 repo id Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Restore the DenseDraft spec class docstrings Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
@WoosukKwon Feel free to merge this after your review.