xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-23 18:25:37 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Bugfix] Allow shared_experts skip quantization for DeepSeekV2/V3 (#14100)

Browse Source 
Signed-off-by: mgoin <mgoin64@gmail.com>
This commit is contained in:
Michael Goin 2025-03-03 16:20:24 -05:00 committed by GitHub
parent ae122b1cbd
commit 2b04c209ee
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 1 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
vllm/model_executor/models/deepseek_v2.py
View File

@ -145,6 +145,7 @@ class DeepseekV2MoE(nn.Module):
hidden_act=config.hidden_act, hidden_act=config.hidden_act,
quant_config=quant_config, quant_config=quant_config,
reduce_results=False, reduce_results=False,
prefix=f"{prefix}.shared_experts",
) )
def forward(self, hidden_states: torch.Tensor) -> torch.Tensor: def forward(self, hidden_states: torch.Tensor) -> torch.Tensor: