EngineX/xc-llm-ascend
3
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

[CI] Fix broken ci (#2530)

Browse Source 
vLLM commit https://github.com/vllm-project/vllm/pull/22711 changed the
encode cache entries logic, this PR adapt the same change for vllm
ascend to make CI happy.

Co-Authored-By: zhoux77899 <zhouxiang100@huawei.com>

- vLLM version: v0.10.1.1
- vLLM main:
0ff902f3b4

Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>
This commit is contained in:
wangxiyuan
2025-08-26 07:42:24 +08:00
committed by GitHub
parent 99bf25af76
commit 7e494e94a9
5 changed files with 256 additions and 123 deletions
Download Patch File Download Diff File Expand all files Collapse all files

55
vllm_ascend/core/scheduler.py
View File

@@ -385,23 +385,44 @@ class AscendScheduler(Scheduler):
req_to_new_blocks)
scheduled_cached_reqs = cached_reqs_data
scheduler_output = SchedulerOutput(
scheduled_new_reqs=new_reqs_data,
scheduled_cached_reqs=scheduled_cached_reqs,
num_scheduled_tokens=num_scheduled_tokens,
total_num_scheduled_tokens=total_num_scheduled_tokens,
scheduled_spec_decode_tokens=scheduled_spec_decode_tokens,
scheduled_encoder_inputs={},
num_common_prefix_blocks=num_common_prefix_blocks,
# finished_req_ids is an existing state in the scheduler,
# instead of being newly scheduled in this step.
# It contains the request IDs that are finished in between
# the previous and the current steps.
finished_req_ids=self.finished_req_ids, # type: ignore
free_encoder_input_ids=self.encoder_cache_manager.get_freed_ids(),
structured_output_request_ids={},
grammar_bitmask=None,
)
if vllm_version_is("0.10.1.1"):
scheduler_output = SchedulerOutput(
scheduled_new_reqs=new_reqs_data,
scheduled_cached_reqs=scheduled_cached_reqs,
num_scheduled_tokens=num_scheduled_tokens,
total_num_scheduled_tokens=total_num_scheduled_tokens,
scheduled_spec_decode_tokens=scheduled_spec_decode_tokens,
scheduled_encoder_inputs={},
num_common_prefix_blocks=num_common_prefix_blocks,
# finished_req_ids is an existing state in the scheduler,
# instead of being newly scheduled in this step.
# It contains the request IDs that are finished in between
# the previous and the current steps.
finished_req_ids=self.finished_req_ids, # type: ignore
free_encoder_input_ids=self.encoder_cache_manager.
get_freed_ids(),
structured_output_request_ids={},
grammar_bitmask=None,
)
else:
scheduler_output = SchedulerOutput(
scheduled_new_reqs=new_reqs_data,
scheduled_cached_reqs=scheduled_cached_reqs,
num_scheduled_tokens=num_scheduled_tokens,
total_num_scheduled_tokens=total_num_scheduled_tokens,
scheduled_spec_decode_tokens=scheduled_spec_decode_tokens,
scheduled_encoder_inputs={},
num_common_prefix_blocks=num_common_prefix_blocks,
# finished_req_ids is an existing state in the scheduler,
# instead of being newly scheduled in this step.
# It contains the request IDs that are finished in between
# the previous and the current steps.
finished_req_ids=self.finished_req_ids, # type: ignore
free_encoder_mm_hashes=self.encoder_cache_manager.
get_freed_mm_hashes(),
structured_output_request_ids={},
grammar_bitmask=None,
)
# NOTE(Kuntai): this function is designed for multiple purposes:
# 1. Plan the KV cache store