Skip to content

Commit a929699

Browse files
[Misc][V0 Deprecation] Remove multi-step worker (#1809)
### What this PR does / why we need it? Remove multi-step worker This PR is a part of #1620. - vLLM version: v0.9.2 - vLLM main: vllm-project/vllm@235bfd5 --------- Signed-off-by: shen-shanshan <467638484@qq.com>
1 parent bf25498 commit a929699

File tree

4 files changed

+0
-303
lines changed

4 files changed

+0
-303
lines changed

vllm_ascend/patch/__init__.py

Lines changed: 0 additions & 17 deletions
Original file line numberDiff line numberDiff line change
@@ -73,23 +73,6 @@
7373
# Future Plan:
7474
# Keep this patch in vllm-ascend.
7575
#
76-
# ** File: worker/patch_common/patch_multi_step_worker.py **
77-
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
78-
# 1. `vllm.spec_decode.multi_step_worker.MultiStepWorker.sampler_output`
79-
# Why:
80-
# There are cuda hard code (current_platform.is_cuda_alike()) in
81-
# `MultiStepWorker.sampler_output`, and we need to use the patched `TP1DraftModelRunner` in it.
82-
# How:
83-
# Make speculative decoding extensible to different backends.
84-
# - support attention metadata register to the set supported spec decode
85-
# - offer a api in platform to determine whether spec decode is supported,
86-
# and deprecate is_cuda_alike in it.
87-
# Related PR (if no, explain why):
88-
# - https://github.com/vllm-project/vllm/pull/15195
89-
# - https://github.com/vllm-project/vllm-ascend/pull/395
90-
# Future Plan:
91-
# Revert it when the related pr is merged in vllm and vllm-ascend.
92-
#
9376
# ** File: worker/patch_common/patch_spec_decode_worker.py **
9477
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
9578
# 1. `vllm.spec_decode.spec_decode_worker.SpecDecodeWorker.create_worker`

vllm_ascend/patch/worker/patch_common/__init__.py

Lines changed: 0 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -20,6 +20,5 @@
2020
import vllm_ascend.patch.worker.patch_common.patch_utils # noqa isort:skip
2121
import vllm_ascend.patch.worker.patch_common.patch_distributed # noqa
2222
import vllm_ascend.patch.worker.patch_common.patch_minicpm # noqa
23-
import vllm_ascend.patch.worker.patch_common.patch_multi_step_worker # noqa
2423
import vllm_ascend.patch.worker.patch_common.patch_sampler # noqa
2524
import vllm_ascend.patch.worker.patch_common.patch_spec_decode_worker # noqa

vllm_ascend/patch/worker/patch_common/patch_multi_step_worker.py

Lines changed: 0 additions & 91 deletions
This file was deleted.

vllm_ascend/worker/multi_step_worker.py

Lines changed: 0 additions & 194 deletions
This file was deleted.

0 commit comments

Comments
 (0)