Skip to content

Add custom kernel for RMS normalization #16

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 4 commits into from
Mar 31, 2023
Merged

Add custom kernel for RMS normalization #16

merged 4 commits into from
Mar 31, 2023

Conversation

WoosukKwon
Copy link
Collaborator

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

@WoosukKwon WoosukKwon requested a review from zhuohan123 March 31, 2023 01:25
Copy link
Member

@zhuohan123 zhuohan123 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

@zhuohan123 zhuohan123 merged commit 09e9245 into main Mar 31, 2023
@WoosukKwon WoosukKwon deleted the rms-norm branch March 31, 2023 16:51
hongxiayang pushed a commit to hongxiayang/vllm that referenced this pull request Feb 13, 2024
AdrianAbeyta added a commit to AdrianAbeyta/vllm that referenced this pull request Mar 8, 2024
Generalizing KV scales JSON to updated schema
luo-cheng2021 pushed a commit to luo-cheng2021/vllm that referenced this pull request Apr 1, 2024
z103cb added a commit to dtrifiro/vllm that referenced this pull request May 8, 2024
fxmarty pushed a commit to fxmarty/vllm-public that referenced this pull request May 31, 2024
yukavio pushed a commit to yukavio/vllm that referenced this pull request Jul 3, 2024
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 <[email protected]>
yukavio pushed a commit to yukavio/vllm that referenced this pull request Jul 3, 2024
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 <[email protected]>
@alixiaodi alixiaodi mentioned this pull request Aug 2, 2024
chaunceyjiang added a commit to chaunceyjiang/vllm that referenced this pull request Apr 23, 2025
robertgshaw2-redhat added a commit to robertgshaw2-redhat/vllm that referenced this pull request May 1, 2025
robertgshaw2-redhat added a commit to robertgshaw2-redhat/vllm that referenced this pull request May 3, 2025
* [Update] LMcache connector v1 implementation

Signed-off-by: ApostaC <[email protected]>

* [Add] examples for disaggregated prefill

Signed-off-by: ApostaC <[email protected]>

* [add] extra information about evns

Signed-off-by: ApostaC <[email protected]>

* Initial stubs for P/D scheduling changes

Signed-off-by: Tyler Michael Smith <[email protected]>

* Updates

Signed-off-by: Tyler Michael Smith <[email protected]>

* Rs branch (#3)

* updated

Signed-off-by: [email protected] <[email protected]>

* Rs branch (#5)

Signed-off-by: [email protected] <[email protected]>

* Remove Unneeded Arguments (#7)

* updated

Signed-off-by: [email protected] <[email protected]>

* stash

Signed-off-by: [email protected] <[email protected]>

* cleanup

Signed-off-by: [email protected] <[email protected]>

---------

Signed-off-by: [email protected] <[email protected]>

* Improve disagg-example.sh (#8)

- fix spelling
- CUDA_VISIBLE_DEVICES should be set externally

Signed-off-by: Tyler Michael Smith <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* added connector

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* update

Signed-off-by: [email protected] <[email protected]>

* remove

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* seems to load properly

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* Revert "updated"

This reverts commit 97316d9.

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* stash

Signed-off-by: [email protected] <[email protected]>

* added

Signed-off-by: [email protected] <[email protected]>

* diffs for local dev on macos

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* update

Signed-off-by: Robert Shaw <[email protected]>

* updaed

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* Checkpoint.

Signed-off-by: Tyler Michael Smith <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* Cleanup

Signed-off-by: Tyler Michael Smith <[email protected]>

* WIP

Signed-off-by: Tyler Michael Smith <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated on scheduler side

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* Hacking away

Signed-off-by: Tyler Michael Smith <[email protected]>

* cleanup

Signed-off-by: Robert Shaw <[email protected]>

* ensure request removed from running list

Signed-off-by: Robert Shaw <[email protected]>

* Runs E2E. Garbage output. Crashes on 2nd request

Signed-off-by: Tyler Michael Smith <[email protected]>

* update

Signed-off-by: Tyler Michael Smith <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* rename files

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* update

Signed-off-by: Robert Shaw <[email protected]>

* Second request no longer crashes

Signed-off-by: Tyler Michael Smith <[email protected]>

* Remove gpu_model_runner hacks

Signed-off-by: Tyler Michael Smith <[email protected]>

* Clean up Justfile

Signed-off-by: Tyler Michael Smith <[email protected]>

* [Bugfix] Stale finished requests in EMPTY_MODEL_RUNNER_OUTPUT

Signed-off-by: Tyler Michael Smith <[email protected]>

* update

Signed-off-by: Tyler Michael Smith <[email protected]>

* justfile edits

Signed-off-by: Tyler Michael Smith <[email protected]>

* Update

Signed-off-by: Tyler Michael Smith <[email protected]>

* Fixes - lm_eval gsm8k has correctness

Signed-off-by: Tyler Michael Smith <[email protected]>

* "just delete the assert"

Signed-off-by: Tyler Michael Smith <[email protected]>

* fixup precommit issues

Signed-off-by: Tyler Michael Smith <[email protected]>

* Fixes

Signed-off-by: Tyler Michael Smith <[email protected]>

* updated (#12)

Signed-off-by: [email protected] <[email protected]>

* Add Accuracy Test (#13)

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

---------

Signed-off-by: [email protected] <[email protected]>

* Preemption Bugfixes (vllm-project#15)

* stash fixed double free issue

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* fixed issue

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updatrd

Signed-off-by: [email protected] <[email protected]>

* updatrd

Signed-off-by: [email protected] <[email protected]>

* updatrd

Signed-off-by: [email protected] <[email protected]>

* updatrd

Signed-off-by: [email protected] <[email protected]>

* updatrd

Signed-off-by: [email protected] <[email protected]>

* updatrd

Signed-off-by: [email protected] <[email protected]>

---------

Signed-off-by: [email protected] <[email protected]>

* updated (vllm-project#16)

Signed-off-by: [email protected] <[email protected]>

* Fix Bad Merge | Fix Memory Leak in Upstream (vllm-project#18)

* updated

Signed-off-by: [email protected] <[email protected]>

* fix merge

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

---------

Signed-off-by: [email protected] <[email protected]>

* clean up justfile, examples

Signed-off-by: Tyler Michael Smith <[email protected]>

* more cleanup

Signed-off-by: Tyler Michael Smith <[email protected]>

* more cleanup

Signed-off-by: Tyler Michael Smith <[email protected]>

* more cleanup

Signed-off-by: Tyler Michael Smith <[email protected]>

* more cleanup

Signed-off-by: Tyler Michael Smith <[email protected]>

* More cleanup

Signed-off-by: Tyler Michael Smith <[email protected]>

* more cleanup

Signed-off-by: Tyler Michael Smith <[email protected]>

* more cleanup, precommit fixes

Signed-off-by: Tyler Michael Smith <[email protected]>

* More cleanup

Signed-off-by: Tyler Michael Smith <[email protected]>

* run_accuracy_test.sh UX

Signed-off-by: Tyler Michael Smith <[email protected]>

* squash warnings

Signed-off-by: Tyler Michael Smith <[email protected]>

* pre-commit

Signed-off-by: Tyler Michael Smith <[email protected]>

* update

Signed-off-by: Tyler Michael Smith <[email protected]>

* Add get_finished to base kv connector

Signed-off-by: mgoin <[email protected]>

* revert test.txt

Signed-off-by: Tyler Michael Smith <[email protected]>

* cleanup

Signed-off-by: Tyler Michael Smith <[email protected]>

* Cleanup

Signed-off-by: Tyler Michael Smith <[email protected]>

* review comments

Signed-off-by: Tyler Michael Smith <[email protected]>

---------

Signed-off-by: ApostaC <[email protected]>
Signed-off-by: Tyler Michael Smith <[email protected]>
Signed-off-by: [email protected] <[email protected]>
Signed-off-by: Robert Shaw <[email protected]>
Signed-off-by: mgoin <[email protected]>
Co-authored-by: ApostaC <[email protected]>
Co-authored-by: Robert Shaw <[email protected]>
Co-authored-by: [email protected] <[email protected]>
Co-authored-by: Robert Shaw <[email protected]>
Co-authored-by: mgoin <[email protected]>
Co-authored-by: mgoin <[email protected]>
robertgshaw2-redhat added a commit to robertgshaw2-redhat/vllm that referenced this pull request May 4, 2025
* [Update] LMcache connector v1 implementation

Signed-off-by: ApostaC <[email protected]>

* [Add] examples for disaggregated prefill

Signed-off-by: ApostaC <[email protected]>

* [add] extra information about evns

Signed-off-by: ApostaC <[email protected]>

* Initial stubs for P/D scheduling changes

Signed-off-by: Tyler Michael Smith <[email protected]>

* Updates

Signed-off-by: Tyler Michael Smith <[email protected]>

* Rs branch (#3)

* updated

Signed-off-by: [email protected] <[email protected]>

* Rs branch (#5)

Signed-off-by: [email protected] <[email protected]>

* Remove Unneeded Arguments (#7)

* updated

Signed-off-by: [email protected] <[email protected]>

* stash

Signed-off-by: [email protected] <[email protected]>

* cleanup

Signed-off-by: [email protected] <[email protected]>

---------

Signed-off-by: [email protected] <[email protected]>

* Improve disagg-example.sh (#8)

- fix spelling
- CUDA_VISIBLE_DEVICES should be set externally

Signed-off-by: Tyler Michael Smith <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* added connector

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* update

Signed-off-by: [email protected] <[email protected]>

* remove

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* seems to load properly

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* Revert "updated"

This reverts commit 97316d9.

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* stash

Signed-off-by: [email protected] <[email protected]>

* added

Signed-off-by: [email protected] <[email protected]>

* diffs for local dev on macos

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* update

Signed-off-by: Robert Shaw <[email protected]>

* updaed

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* Checkpoint.

Signed-off-by: Tyler Michael Smith <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* Cleanup

Signed-off-by: Tyler Michael Smith <[email protected]>

* WIP

Signed-off-by: Tyler Michael Smith <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated on scheduler side

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* Hacking away

Signed-off-by: Tyler Michael Smith <[email protected]>

* cleanup

Signed-off-by: Robert Shaw <[email protected]>

* ensure request removed from running list

Signed-off-by: Robert Shaw <[email protected]>

* Runs E2E. Garbage output. Crashes on 2nd request

Signed-off-by: Tyler Michael Smith <[email protected]>

* update

Signed-off-by: Tyler Michael Smith <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* rename files

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* update

Signed-off-by: Robert Shaw <[email protected]>

* Second request no longer crashes

Signed-off-by: Tyler Michael Smith <[email protected]>

* Remove gpu_model_runner hacks

Signed-off-by: Tyler Michael Smith <[email protected]>

* Clean up Justfile

Signed-off-by: Tyler Michael Smith <[email protected]>

* [Bugfix] Stale finished requests in EMPTY_MODEL_RUNNER_OUTPUT

Signed-off-by: Tyler Michael Smith <[email protected]>

* update

Signed-off-by: Tyler Michael Smith <[email protected]>

* justfile edits

Signed-off-by: Tyler Michael Smith <[email protected]>

* Update

Signed-off-by: Tyler Michael Smith <[email protected]>

* Fixes - lm_eval gsm8k has correctness

Signed-off-by: Tyler Michael Smith <[email protected]>

* "just delete the assert"

Signed-off-by: Tyler Michael Smith <[email protected]>

* fixup precommit issues

Signed-off-by: Tyler Michael Smith <[email protected]>

* Fixes

Signed-off-by: Tyler Michael Smith <[email protected]>

* updated (#12)

Signed-off-by: [email protected] <[email protected]>

* Add Accuracy Test (#13)

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

---------

Signed-off-by: [email protected] <[email protected]>

* Preemption Bugfixes (vllm-project#15)

* stash fixed double free issue

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* fixed issue

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updatrd

Signed-off-by: [email protected] <[email protected]>

* updatrd

Signed-off-by: [email protected] <[email protected]>

* updatrd

Signed-off-by: [email protected] <[email protected]>

* updatrd

Signed-off-by: [email protected] <[email protected]>

* updatrd

Signed-off-by: [email protected] <[email protected]>

* updatrd

Signed-off-by: [email protected] <[email protected]>

---------

Signed-off-by: [email protected] <[email protected]>

* updated (vllm-project#16)

Signed-off-by: [email protected] <[email protected]>

* Fix Bad Merge | Fix Memory Leak in Upstream (vllm-project#18)

* updated

Signed-off-by: [email protected] <[email protected]>

* fix merge

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

---------

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* cleanup code

Signed-off-by: [email protected] <[email protected]>

* cleanup code

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* stash

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updatted

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* revert

Signed-off-by: [email protected] <[email protected]>

* more spurious changes

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* Update vllm/distributed/kv_transfer/kv_connector/v1/nixl_connector.py

Co-authored-by: Tyler Michael Smith <[email protected]>

* Update vllm/distributed/kv_transfer/kv_connector/v1/nixl_connector.py

Co-authored-by: Tyler Michael Smith <[email protected]>

---------

Signed-off-by: ApostaC <[email protected]>
Signed-off-by: Tyler Michael Smith <[email protected]>
Signed-off-by: [email protected] <[email protected]>
Signed-off-by: Robert Shaw <[email protected]>
Co-authored-by: ApostaC <[email protected]>
Co-authored-by: Tyler Michael Smith <[email protected]>
Co-authored-by: Tyler Michael Smith <[email protected]>
Co-authored-by: Robert Shaw <[email protected]>
CatherineSue added a commit to moirai-internal/vllm that referenced this pull request May 5, 2025
Merge in GEN/vllm from chang/embedding to embedding-based-on-v0.3.3.0

Squashed commit of the following:

commit d5ac491c0b506f25c0e2576e00d5abcb839bea21
Author: Chang Su <[email protected]>
Date:   Tue Mar 19 18:08:55 2024 -0700

    Add validation for max_tokens in EmbeddingRequest

commit 0cb9a42326af8b5b0a12b8627f3ee0d8097ed7e2
Author: Chang Su <[email protected]>
Date:   Tue Mar 19 17:54:15 2024 -0700

    Fix __repr__ of EmbeddingSequenceGroupOutput

    - List type doesn't have shape attr

commit 9df0f9d70d871d015867941d0c9677345b6d642f
Author: Chang Su <[email protected]>
Date:   Tue Mar 19 17:21:52 2024 -0700

    Update embeddings type

    - Update embeddings type to List[float]
    - Explicitly cast torch.Tensor to list
    - Remove usused imports

commit 5946fa8eb8ba134e868d52637987c383088d39d4
Author: Chang Su <[email protected]>
Date:   Tue Mar 19 16:33:19 2024 -0700

    Resolve comments

    - Remove unused imports in llama.py
    - Change embedding type to torch.Tensor

commit 529911bc8708c79a382c1c803d158d3818b33cb0
Author: Chang Su <[email protected]>
Date:   Tue Mar 19 13:45:07 2024 -0700

    Add Embedding API to entrypoints/openai

    - Add EmbeddingRequest and Response to protocol
    - Add serving_embedding
    - Add OpenAIServingEmbedding to api_server

commit cb09c286e1910fce1e98d92c17140f874b95a6f7
Author: Chang Su <[email protected]>
Date:   Tue Mar 19 13:41:23 2024 -0700

    Add Embedding in outputs and sequence

    - Add EmbeddingOutput, EmbeddingRequestOutput, RequestOutputFactory and
    EmbeddingSequenceGroupOutput to support process the embedding output
    sequence
    - Update process output and sequence in *llm_engine to use embedding

commit 5cd2c198c3082f08f9a2545c01c547908b031acc
Author: Chang Su <[email protected]>
Date:   Tue Mar 19 13:38:21 2024 -0700

    Add MistralModel

    - Add embedding and load_weights in LlamaModel to support forward
    - Mistral uses LlamaModel
    - Use embedding when embedding_mode is True in embedding_model_runner
robertgshaw2-redhat added a commit to robertgshaw2-redhat/vllm that referenced this pull request May 6, 2025
* [Update] LMcache connector v1 implementation

Signed-off-by: ApostaC <[email protected]>

* [Add] examples for disaggregated prefill

Signed-off-by: ApostaC <[email protected]>

* [add] extra information about evns

Signed-off-by: ApostaC <[email protected]>

* Initial stubs for P/D scheduling changes

Signed-off-by: Tyler Michael Smith <[email protected]>

* Updates

Signed-off-by: Tyler Michael Smith <[email protected]>

* Rs branch (#3)

* updated

Signed-off-by: [email protected] <[email protected]>

* Rs branch (#5)

Signed-off-by: [email protected] <[email protected]>

* Remove Unneeded Arguments (#7)

* updated

Signed-off-by: [email protected] <[email protected]>

* stash

Signed-off-by: [email protected] <[email protected]>

* cleanup

Signed-off-by: [email protected] <[email protected]>

---------

Signed-off-by: [email protected] <[email protected]>

* Improve disagg-example.sh (#8)

- fix spelling
- CUDA_VISIBLE_DEVICES should be set externally

Signed-off-by: Tyler Michael Smith <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* added connector

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* update

Signed-off-by: [email protected] <[email protected]>

* remove

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* seems to load properly

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* Revert "updated"

This reverts commit 97316d9.

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* stash

Signed-off-by: [email protected] <[email protected]>

* added

Signed-off-by: [email protected] <[email protected]>

* diffs for local dev on macos

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* update

Signed-off-by: Robert Shaw <[email protected]>

* updaed

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* Checkpoint.

Signed-off-by: Tyler Michael Smith <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* Cleanup

Signed-off-by: Tyler Michael Smith <[email protected]>

* WIP

Signed-off-by: Tyler Michael Smith <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated on scheduler side

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* Hacking away

Signed-off-by: Tyler Michael Smith <[email protected]>

* cleanup

Signed-off-by: Robert Shaw <[email protected]>

* ensure request removed from running list

Signed-off-by: Robert Shaw <[email protected]>

* Runs E2E. Garbage output. Crashes on 2nd request

Signed-off-by: Tyler Michael Smith <[email protected]>

* update

Signed-off-by: Tyler Michael Smith <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* rename files

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* updated

Signed-off-by: Robert Shaw <[email protected]>

* update

Signed-off-by: Robert Shaw <[email protected]>

* Second request no longer crashes

Signed-off-by: Tyler Michael Smith <[email protected]>

* Remove gpu_model_runner hacks

Signed-off-by: Tyler Michael Smith <[email protected]>

* Clean up Justfile

Signed-off-by: Tyler Michael Smith <[email protected]>

* [Bugfix] Stale finished requests in EMPTY_MODEL_RUNNER_OUTPUT

Signed-off-by: Tyler Michael Smith <[email protected]>

* update

Signed-off-by: Tyler Michael Smith <[email protected]>

* justfile edits

Signed-off-by: Tyler Michael Smith <[email protected]>

* Update

Signed-off-by: Tyler Michael Smith <[email protected]>

* Fixes - lm_eval gsm8k has correctness

Signed-off-by: Tyler Michael Smith <[email protected]>

* "just delete the assert"

Signed-off-by: Tyler Michael Smith <[email protected]>

* fixup precommit issues

Signed-off-by: Tyler Michael Smith <[email protected]>

* Fixes

Signed-off-by: Tyler Michael Smith <[email protected]>

* updated (#12)

Signed-off-by: [email protected] <[email protected]>

* Add Accuracy Test (#13)

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

---------

Signed-off-by: [email protected] <[email protected]>

* Preemption Bugfixes (vllm-project#15)

* stash fixed double free issue

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* fixed issue

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updatrd

Signed-off-by: [email protected] <[email protected]>

* updatrd

Signed-off-by: [email protected] <[email protected]>

* updatrd

Signed-off-by: [email protected] <[email protected]>

* updatrd

Signed-off-by: [email protected] <[email protected]>

* updatrd

Signed-off-by: [email protected] <[email protected]>

* updatrd

Signed-off-by: [email protected] <[email protected]>

---------

Signed-off-by: [email protected] <[email protected]>

* updated (vllm-project#16)

Signed-off-by: [email protected] <[email protected]>

* Fix Bad Merge | Fix Memory Leak in Upstream (vllm-project#18)

* updated

Signed-off-by: [email protected] <[email protected]>

* fix merge

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

---------

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* cleanup code

Signed-off-by: [email protected] <[email protected]>

* cleanup code

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* stash

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updatted

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* revert

Signed-off-by: [email protected] <[email protected]>

* more spurious changes

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* updated

Signed-off-by: [email protected] <[email protected]>

* Support MLA in NIXL connector

Signed-off-by: Tyler Michael Smith <[email protected]>

* WIP adding tests

Signed-off-by: Tyler Michael Smith <[email protected]>

* wip

Signed-off-by: Tyler Michael Smith <[email protected]>

* Fixes

Signed-off-by: Tyler Michael Smith <[email protected]>

---------

Signed-off-by: ApostaC <[email protected]>
Signed-off-by: Tyler Michael Smith <[email protected]>
Signed-off-by: [email protected] <[email protected]>
Signed-off-by: Robert Shaw <[email protected]>
Co-authored-by: ApostaC <[email protected]>
Co-authored-by: Robert Shaw <[email protected]>
Co-authored-by: [email protected] <[email protected]>
Co-authored-by: Robert Shaw <[email protected]>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants