Ruri 866f5e7283 [Bugfix] Fix weight prefetching AssertionError in W8A8 MTP scene (#3361) ...

### What this PR does / why we need it?

- Fix `AssertionError` of `weight_prefetch_method` in W8A8 MTP scene
- Remove hard-code key
(https://github.com/vllm-project/vllm-ascend/pull/3146#discussion_r2416644010)

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

None

### How was this patch tested?
`weight_prefetch_method is None` (tested on DeepSeek-R1-w8a8mix_MTP)

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

Signed-off-by: zhoux77899 <zhouxiang100@huawei.com>

2025-10-11 09:24:02 +08:00

..

__init__.py

[Core] Cherry pick from 0.7.1 to keep the main code newest (#127)

2025-02-21 17:07:37 +08:00

quant_config.py

[MoE] [Refactor] Combine common_fused_moe and fused_moe (#3176)

2025-10-09 14:12:46 +08:00

utils.py

[Quantization] register AscendQuantRMSNorm for quantization (#2856)

2025-09-11 23:14:02 +08:00

w4a8_dynamic.py

[main][quantization] Support deepseek w4a8 per-channel quantization (#3011)

2025-09-27 21:01:16 +08:00

w8a8_dynamic.py

[Feat] Load balance of tokens across experts in dummy_run (#3184)

2025-10-10 09:00:07 +08:00

w8a8.py

[Bugfix] Fix weight prefetching AssertionError in W8A8 MTP scene (#3361)

2025-10-11 09:24:02 +08:00