LookAround0301 f8610b7d67 [long_seq] fix A2 accuracy problem (#4030) ...

### What this PR does / why we need it?
1、update prepare_finalize.py：fix A2 accuracy problem when pcp and dcp

- vLLM version: v0.11.0
- vLLM main:
83f478bb19

Signed-off-by: LookAround <lixushi@huawei.com>

2025-11-07 09:29:33 +08:00

..

attention

[BugFix][main] Adapted to torch_npu.npu_fused_infer_attention_score (#4025)

2025-11-06 22:00:24 +08:00

compilation

[main][bugfix] Fix a rare bug triggered by _npu_paged_attention in FULL_DECODE_ONLY mode (#3986)

2025-11-06 23:08:07 +08:00

core

Upgrade to new vllm commit (#3719)

2025-10-25 15:36:32 +08:00

device_allocator

[Misc]Clean up useless import from vllm (#2049)

2025-07-28 16:01:59 +08:00

distributed

[Feat](Mooncake) Supports multiple input suffixes for global_segment_size (#3690)

2025-11-06 14:48:15 +08:00

eplb

[CI]Fix oom of deepseek-eplb nigtly test. (#3884)

2025-10-30 10:18:07 +08:00

kv_offload

Upgrade to 0.11.1 newest vllm commit (#3762)

2025-10-28 14:55:03 +08:00

lora

[1/N][Refactor] Refactor code to adapt with vllm main (#3612)

2025-10-24 16:55:08 +08:00

model_loader

Upgrade to new vllm commit (#3719)

2025-10-25 15:36:32 +08:00

models

[Model][3/N] Refactor sfa into mla and remove deepseek_v3_2.py (#3769)

2025-10-30 17:06:38 +08:00

ops

[long_seq] fix A2 accuracy problem (#4030)

2025-11-07 09:29:33 +08:00

patch

[BugFix][main] Adapted to torch_npu.npu_fused_infer_attention_score (#4025)

2025-11-06 22:00:24 +08:00

quantization

[Perf] move quant before allgather in Allgather EP (#3420)

2025-11-04 16:49:58 +08:00

sample

Upgrade to 0.11.1 newest vllm commit (#3762)

2025-10-28 14:55:03 +08:00

spec_decode

[BugFix] Fix deepseek v3.2 mtp bug. (#3900)

2025-11-04 14:06:59 +08:00

torchair

support qwen3-next full_decode_only mode. (#3949)

2025-11-05 08:46:05 +08:00

worker

[feat]decode convert bsnd to tnd and fix bug when pcp and dcp (#3980)

2025-11-06 14:58:24 +08:00

__init__.py

[1/N][Refactor] Refactor code to adapt with vllm main (#3612)

2025-10-24 16:55:08 +08:00

ascend_config.py

[1/N][Refactor] Refactor code to adapt with vllm main (#3612)

2025-10-24 16:55:08 +08:00

ascend_forward_context.py

[Bugfix] Add constraints for sequence parallelism (#4014)

2025-11-06 20:02:03 +08:00

cpu_binding.py

[main] support cpu binding (#3546)

2025-10-21 09:17:03 +08:00

envs.py

[main] remove dbo code (#3712)

2025-10-25 15:53:01 +08:00

meta_registration.py

Fix the bugs about operator registration by PyTorch Dispatcher (#2786)

2025-09-13 11:58:52 +08:00

platform.py

mfix bug when max_seqs=14 in mtp=2 scenario and raise error when cudagraph_capture_sizes can't be an integer multiple of uniform_decode_query_lentp (#3910)

2025-10-31 09:24:50 +08:00

utils.py

[Bugfix] Add constraints for sequence parallelism (#4014)

2025-11-06 20:02:03 +08:00