vlm: support video as an input modality (#5888)
This commit is contained in:
@@ -1827,8 +1827,7 @@ class MiniCPMO(MiniCPMBaseModel):
|
||||
input_ids=input_ids,
|
||||
forward_batch=forward_batch,
|
||||
language_model=self.llm,
|
||||
image_data_embedding_func=self.get_image_feature,
|
||||
audio_data_embedding_func=self.get_audio_feature,
|
||||
multimodal_model=self,
|
||||
positions=positions,
|
||||
)
|
||||
return hidden_states
|
||||
|
||||
Reference in New Issue
Block a user