Open
Description
🚀 The feature, motivation and pitch
After 0.9.2rc1, v0 code will be removed from vllm-ascend. This issue list the todo list.
Prepare
- Enable V1 by default follow: vllm-project/vllm@5a7f6a7
- Release v0.9.2rc1
todo list
- Cleanup all docs and test related to V0 [Test] Remove VLLM_USE_V1 in example and tests #1733
- Cleanup VLLM_USE_V1 in function code [Misc] Remove VLLM_USE_V1 usage in code #1764
- Cleanup all benchmark, accuracy test, long-term test, pd test, doc test CI V0 code [Misc][V0 Deprecation] Remove V0 Related Codes of Test, Example and Platform #1805
- Cleanup all code related to V0 in function code @shen-shanshan
- Cleanup v0 worker [Misc][V0 Deprecation] Remove V0 Worker #1821
- Cleanup v0 model_runner [Misc][V0 Deprecation] Remove V0 Model Runner #1823
- Cleanup multi step worker [Misc][V0 Deprecation] Remove Multi-Step Worker #1809
- Cleanup multi step model_runner [Misc][V0 Deprecation] Remove Multi-Step Model Runner #1820
- Cleanup draft_model_runner [Misc][V0 Deprecation] Remove Draft Model Runner Used for V0 Spec Decode #1810
- Cleanup pooling_model_runner [Misc][V0 Deprecation] Remove Pooling Model Runner #1824
- Cleanup v0 attention
- Cleanup v0 patch
- Cleanup V0 PD code
- Cleanup V0 custom ops
- Rename all v1 file, for example rename
worker_v1toworker. - Rename related code following vllm, for example
prompt_adapter
Alternatives
No response
Additional context
No response