zzzzwwjj f1543d5e0d [bugfix] fix deeepseek accuracy (#1118) ...

### What this PR does / why we need it?
fix deeepseek accuracy in mix-parallel case.


Signed-off-by: zzzzwwjj <1183291235@qq.com>

2025-06-07 21:11:36 +08:00

..

attention

[perf]: support dual-batch overlap(dbo) for deepseek (#941)

2025-06-07 16:46:58 +08:00

compilation

[aclgraph] implentment NPUPiecewiseBackend to enable aclgraph (#836)

2025-05-29 11:58:26 +08:00

core

[Misc] fix initialize_kv_cache (#1102)

2025-06-06 16:46:23 +08:00

device_allocator

[bugfix] Improve log level and info for custom ops build (#937)

2025-05-23 10:05:57 +08:00

distributed

[bugfix] some bugs maybe fail to run (#896)

2025-06-03 11:07:33 +08:00

lora

[Bugfix] fix import error (#600)

2025-04-22 08:57:25 +08:00

models

[bugfix] fix deeepseek accuracy (#1118)

2025-06-07 21:11:36 +08:00

multistream

[perf]: support dual-batch overlap(dbo) for deepseek (#941)

2025-06-07 16:46:58 +08:00

ops

[bugfix] fix deeepseek accuracy (#1118)

2025-06-07 21:11:36 +08:00

patch

Fix typo of VLLM_ASCEND_ENABLE_TOPK_OPTIMIZE (#1112)

2025-06-07 19:45:33 +08:00

quantization

[perf]Support MOE Multi-stream in Deepseek (#947)

2025-06-05 23:39:38 +08:00

sample

Spec decode support for V1 Engine (#874)

2025-05-23 14:25:46 +08:00

worker

Add qwen2.5 vl multimodal feature for vllm-ascend v1 (#736)

2025-06-07 16:53:19 +08:00

__init__.py

[CI] Patch torch.library.infer_schema for fused moe ops to fix CI (#854)

2025-05-14 19:49:09 +08:00

ascend_config.py

Avoid unfused Transpose in DeepSeekV3 EP256 MoE layer (#1091)

2025-06-07 14:28:20 +08:00

envs.py

[CherryPick] Add unpadded Qwen2.5-VL for verl scenario (#1095)

2025-06-07 19:45:46 +08:00

platform.py

[bugfix] fix deeepseek accuracy (#1118)

2025-06-07 21:11:36 +08:00

utils.py

[ModelRunner]Add profile execute duration observation (#1013)

2025-06-06 09:29:34 +08:00