sherie f86596a66c allgather use fusedop. (#2689) ...

### What this PR does / why we need it?
Use 'npu_moe_init_routing_v2' &'npu_moe_token_unpermute' repalce
'npu_moe_init_routing' &‘npu_moe_compute_expert_tokens’&
'npu_moe_finalize_routing' to optimize performance
### Does this PR introduce _any_ user-facing change?
| branch| tps| TTFT |TPOT |
| --- | --- | --- |--- |
|main  |733.98  | 280.05 |34.30 |
|main+fusedop  | 740.33 | 273.34 |33.99 |
### How was this patch tested?


- vLLM version: v0.10.1.1
- vLLM main:
6997a25ac6

Signed-off-by: wangxiaoxin-sherie <wangxiaoxin7@huawei.com>
Co-authored-by: wangxiaoxin-sherie <wangxiaoxin7@huawei.com>

2025-09-04 11:56:29 +08:00

..

310p

Refactor e2e CI (#2276)

2025-09-02 09:02:22 +08:00

doctests

Remove transformer pins for v0.9.1-dev (#2234)

2025-08-07 14:41:10 +08:00

models

Accuracy report formatting (#2279)

2025-08-25 09:39:30 +08:00

multicard

[1/N][Draft][Refactor]torchair pangu_moe modeling refactor (#2437)

2025-09-04 10:39:21 +08:00

pd_disaggreate

Disaggregate prefill for kv cache register style (#950)

2025-07-26 17:15:47 +08:00

prompts

[CI/UT] Add test for chunk prefill and prefix cache on v1/AscendScheduler (#1505)

2025-07-02 16:57:03 +08:00

singlecard

allgather use fusedop. (#2689)

2025-09-04 11:56:29 +08:00

__init__.py

[Test] Clean up duplicate test for ascend scheduler (#1819)

2025-07-16 17:57:48 +08:00

common.sh

[Doctest] add installation doctest (#1179)

2025-06-17 08:52:26 +08:00

conftest.py

[CI] recover e2e test (#2688)

2025-09-02 18:49:17 +08:00

model_utils.py

Refactor e2e CI (#2276)

2025-09-02 09:02:22 +08:00

run_disagg_pd.sh

[CI] Fix PD job (#1129)

2025-06-09 16:34:41 +08:00

run_doctests.sh

[Doctest] add installation doctest (#1179)

2025-06-17 08:52:26 +08:00

utils.py

[Test] Remove VLLM_USE_V1 in example and tests (#1733)

2025-07-15 12:49:57 +08:00