[MTP] Refactor mtp predictor to avoid d2h operation (#27643)

Signed-off-by: MengqingCao <[email protected]>

Author: MengqingCao

vllm/model_executor/models/deepseek_mtp.py

@@ -97,7 +97,7 @@ def forward(
     ) -> torch.Tensor:
         assert inputs_embeds is not None
         # masking inputs at position 0, as not needed by MTP
-        inputs_embeds[positions == 0] = 0
+        inputs_embeds = torch.where(positions.unsqueeze(-1) == 0, 0, inputs_embeds)
         inputs_embeds = self.enorm(inputs_embeds)
         previous_hidden_states = self.hnorm(previous_hidden_states)