biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Bugfix] Fix DeepSeek MTP (#22934)

Browse Source 
Signed-off-by: Benjamin Chislett <benjamin.chislett@centml.ai>
This commit is contained in:
Benjamin Chislett
2025-08-15 21:25:06 -04:00
committed by GitHub
parent 070da660c1
commit fbd88728b3
3 changed files with 13 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
vllm/model_executor/models/glm4_moe_mtp.py
View File

@@ -180,14 +180,13 @@ class Glm4MoeMTP(nn.Module, SupportsPP):
self,
input_ids: torch.Tensor,
positions: torch.Tensor,
previous_hidden_states: torch.Tensor,
hidden_states: torch.Tensor,
intermediate_tensors: Optional[IntermediateTensors] = None,
inputs_embeds: Optional[torch.Tensor] = None,
spec_step_idx: int = 0,
) -> torch.Tensor:
hidden_states = self.model(input_ids, positions,
previous_hidden_states, inputs_embeds,
spec_step_idx)
hidden_states = self.model(input_ids, positions, hidden_states,
inputs_embeds, spec_step_idx)
return hidden_states
def compute_logits(