Update modeling_minicpmv.py

modeling_minicpmv.py

@@ -272,7 +272,7 @@ class MiniCPMV(MiniCPMVPreTrainedModel):
             counter += 1
             batch.append(input_embeds)
             
-        batch = torch.stack(batch)
+        batch = torch.cat(batch, dim=0)
         # output_ids = self._decode(input_embeds, tokenizer, **kwargs)
         if stream:
             kwargs.pop("decode_text")