Angazenn d9ee491f70 [BugFix]Move to_list in foward_v1 with FIA earlier to build (#3185) ...

### What this PR does / why we need it?
The current implementation of FIA will introduce an `to_list` operation
for actual_seq_lengths_q and seq_lens，which comsumes extra time. These
operation can be moved earlier into `build` operation of attention
metadata.

### Does this PR introduce _any_ user-facing change?

No.

### How was this patch tested?


- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0

---------

Signed-off-by: Angazenn <supperccell@163.com>

2025-10-17 11:19:41 +08:00

..

__init__.py

[Core] Make V1 work and enable V1 engine test (#389)

2025-03-28 19:34:23 +08:00

attention_mask.py

[Feat] Supports Aclgraph for bge-m3 (#3171)

2025-10-14 23:07:45 +08:00

attention_v1.py

[BugFix]Move to_list in foward_v1 with FIA earlier to build (#3185)

2025-10-17 11:19:41 +08:00

mla_v1.py

[Feat] Flash comm allgher ep (#3334)

2025-10-15 19:36:32 +08:00

sfa_v1.py

[Refactor] Adapt deepseek-v3.2 to vllm 0.11.0 (#3432)

2025-10-15 17:48:58 +08:00

utils.py

[Feat]Qwen3 Moe supports npu_add_rms_norm_quant op by default, update op with bias, resolve conflict with weight prefetch (#3465)

2025-10-17 09:30:51 +08:00