[Core]Init vllm-ascend #2
Closed
Conversation
This file contains 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
Signed-off-by: MengqingCao <[email protected]>
* add requirements.txt for npu * update setup.py Signed-off-by: MengqingCao <[email protected]>
Add npu worker and model_runner
Signed-off-by: MengqingCao <[email protected]>
Signed-off-by: MengqingCao <[email protected]>
Signed-off-by: MengqingCao <[email protected]>
* add block_size arg to inference script * add get_current_memory_usage to platform Signed-off-by: MengqingCao <[email protected]>
Signed-off-by: MengqingCao <[email protected]>
Signed-off-by: MengqingCao <[email protected]>
Signed-off-by: MengqingCao <[email protected]>
[bugfix]: attention alibi bias shape is wrong
[Docs] Update docs and code format
[Format] Update docs and add format.sh
Signed-off-by: MengqingCao <[email protected]>
Signed-off-by: Yikun <[email protected]>
Signed-off-by: Yikun <[email protected]>
Signed-off-by: MengqingCao <[email protected]>
Signed-off-by: MengqingCao <[email protected]>
Signed-off-by: MengqingCao <[email protected]>
Signed-off-by: Shanshan Shen <[email protected]>
Signed-off-by: Yikun <[email protected]>
Signed-off-by: Yikun <[email protected]>
* enable npu profiling Signed-off-by: wangli <[email protected]> * adjust import position Signed-off-by: wangli <[email protected]> --------- Signed-off-by: wangli <[email protected]>
* [CI] use cached vLLM Signed-off-by: MengqingCao <[email protected]> * fix path Signed-off-by: MengqingCao <[email protected]> --------- Signed-off-by: MengqingCao <[email protected]>
Signed-off-by: Yikun <[email protected]>
* update contributing doc Signed-off-by: Shanshan Shen <[email protected]> * update contributing doc Signed-off-by: Shanshan Shen <[email protected]> * update contributing doc Signed-off-by: Shanshan Shen <[email protected]> * update contributing doc Signed-off-by: Shanshan Shen <[email protected]> * add ut pr label Signed-off-by: Shanshan Shen <[email protected]> * update pr labels Signed-off-by: Shanshan Shen <[email protected]> * update pr labels Signed-off-by: Shanshan Shen <[email protected]> --------- Signed-off-by: Shanshan Shen <[email protected]>
Signed-off-by: Yikun <[email protected]>
* Init vLLM Ascend backend plugin README.md Signed-off-by: Yikun <[email protected]> * Fix words and logos Signed-off-by: Yikun <[email protected]> --------- Signed-off-by: Yikun <[email protected]>
### What this PR does / why we need it? Add Python ignore and `.DS_Store` ### Does this PR introduce _any_ user-facing change? No ### How was this patch tested? git status see related file ignored Signed-off-by: Yikun <[email protected]>
### What this PR does / why we need it? This patch align with the meta data of the vllm community and complete the necessary information: - Add copyright - Update version: 0.1.0a1 alpha release follow https://packaging.python.org/en/latest/specifications/version-specifiers - Add author/license/description - Remove extras_require ### Does this PR introduce _any_ user-facing change? No ### How was this patch tested? ``` # Prepare $ python3 -m venv ./.venv-test $ source ./.venv-test/bin/activate # Install without failaure $ pip install -e . --no-deps # Check metadata as exptected $ ls ./.venv-test/lib/python3.12/site-packages/vllm_ascend-0.1.0a1.dist-info INSTALLER LICENSE METADATA RECORD REQUESTED WHEEL direct_url.json entry_points.txt top_level.txt ``` Signed-off-by: Yikun <[email protected]>
init vLLM native UTs Signed-off-by: MengqingCao <[email protected]>
* Refactor worker to decouple it from gpuworker * add `get_model` func to `model_runner` and `worker` plz merge this pr before #32 , because it relys on the reconstruction in this PR Signed-off-by: MengqingCao <[email protected]>
Mainly changes: * exit if vLLM native ut failed * rm redundant mappings in docker * update TEST_FILES `test_config.py` is disabled beacuse some models require for access:  Signed-off-by: MengqingCao <[email protected]>
### What this PR does / why we need it? Adapts AttentionLayer and kvcache scaling interface to make ci happy with master vLLM. - vllm-project/vllm@e97f802 Add the dynamic kv cache scaling factors - vllm-project/vllm@86bfb6d Add AttentionLayer to forward ### Does this PR introduce _any_ user-facing change? No ### How was this patch tested? - run with master vllm ``` # Install vLLM main git clone --depth 1 https://github.com/vllm-project/vllm.git /workspace/vllm python3 -m pip install -r /workspace/vllm/requirements-build.txt VLLM_TARGET_DEVICE="empty" python3 -m pip install /workspace/vllm/ # Install vllm-ascend main python3 -m pip install /workspace/vllm_ascend/ # test pytest tests -sv ``` - CI passed Signed-off-by: Yikun <[email protected]>
### What this PR does / why we need it? Test with vllm master code to fix CI failure ### Does this PR introduce _any_ user-facing change? No ### How was this patch tested? By CI test Signed-off-by: wangxiyuan <[email protected]>
### What this PR does / why we need it? Remove useless ci deps because we use `VLLM_TARGET_DEVICE=empty` to build ### Does this PR introduce _any_ user-facing change? No ### How was this patch tested? CI passed Signed-off-by: Yikun <[email protected]>
### What this PR does / why we need it? - Add `Building and testing` section to CONTRIBUTING.md - Refresh Dockerfile to use empty VLLM_TARGET_DEVICE ### Does this PR introduce _any_ user-facing change? No ### How was this patch tested? Preview and test Signed-off-by: Yikun Jiang <[email protected]>
|
Please Do not merge, the CI need work first. |
Yikun
reviewed
Jan 30, 2025
Yikun
force-pushed
the
main
branch
2 times, most recently
from
83e76e3 to
f782b51
Compare
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.
No description provided.