EngineX/xc-llm-ascend
3
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

[MISC] fix format check error (#654)

Browse Source 
This pr makes format.sh works as expect.

Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>
This commit is contained in:
wangxiyuan
2025-04-29 11:14:19 +08:00
committed by GitHub
parent 1fce70a2fb
commit 0dae55a9a3
17 changed files with 45 additions and 44 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
vllm_ascend/ops/fused_moe.py
View File

@@ -278,7 +278,7 @@ def fused_experts(
dtype=dtype)
# TODO: npu_grouped_matmul output random values at [num_valid_tokens:, ...]
# This created multiple NaN and index_add_ will mix them up which harms accracy
# This created multiple NaN and index_add_ will mix them up which harms accuracy
# remove this mask and filter after it being fixed
num_valid_tokens = mask.sum()
valid_token_mask = torch.arange(

2
vllm_ascend/ops/rotary_embedding.py
View File

@@ -227,7 +227,7 @@ def _set_cos_sin_cache(self, seq_len, device, dtype):
persistent=False)
# TODO: Patch when aclnn ops avaiable
# TODO: Patch when aclnn ops available
RotaryEmbedding.forward_oot = rope_forward_oot
DeepseekScalingRotaryEmbedding.forward = native_rope_deepseek_forward
DeepseekScalingRotaryEmbedding._set_cos_sin_cache = _set_cos_sin_cache