EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity

Tiny fix wrong comments (#9589)

Browse Source
This commit is contained in:
fzyzcjy
2025-08-25 18:08:10 +08:00
committed by GitHub
parent f8b757bcac
commit 9dcdf5da03
1 changed files with 0 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
python/sglang/srt/layers/quantization/modelopt_quant.py
View File

@@ -876,7 +876,6 @@ class ModelOptNvFp4FusedMoEMethod(FusedMoEMethodBase):
data=torch.empty( data=torch.empty(
layer.num_local_experts, layer.num_local_experts,
2 * intermediate_size_per_partition, 2 * intermediate_size_per_partition,
# 2 fp4 items are packed in the input dimension
hidden_size // self.quant_config.group_size, hidden_size // self.quant_config.group_size,
dtype=weight_scale_dtype, dtype=weight_scale_dtype,
), ),
@@ -895,7 +894,6 @@ class ModelOptNvFp4FusedMoEMethod(FusedMoEMethodBase):
data=torch.empty( data=torch.empty(
layer.num_local_experts, layer.num_local_experts,
hidden_size, hidden_size,
# 2 fp4 items are packed in the input dimension
intermediate_size_per_partition // self.quant_config.group_size, intermediate_size_per_partition // self.quant_config.group_size,
dtype=weight_scale_dtype, dtype=weight_scale_dtype,
), ),