[Bugfix] Fix OpenPangu multimodal embedding merge (#55941)

Signed-off-by: Taneem Ibrahim <taneem.ibrahim@gmail.com>
This commit is contained in:
Taneem Ibrahim
2026-09-08 23:53:59 +00:00
committed by GitHub
parent 9c2d21046b
commit 5af4cc33ec
2 changed files with 68 additions and 10 deletions
+13 -10
View File
@@ -1039,17 +1039,20 @@ class OpenPanguVLForConditionalGeneration(
def get_input_embeddings(
self,
input_ids: torch.Tensor,
multimodal_embeddings=None,
multimodal_embeddings: MultiModalEmbeddings | None = None,
) -> torch.Tensor:
inputs_embeds = self.language_model.embed_input_ids(input_ids)
if multimodal_embeddings is not None:
inputs_embeds = self.embed_input_ids(
input_ids,
inputs_embeds,
multimodal_embeddings,
[self.config.image_token_id, self.config.video_token_id],
)
return inputs_embeds
if multimodal_embeddings is None:
return self.language_model.embed_input_ids(input_ids)
mm_token_ids = input_ids.new_tensor(
[self.config.image_token_id, self.config.video_token_id]
)
is_multimodal = torch.isin(input_ids, mm_token_ids)
return self.embed_input_ids(
input_ids,
multimodal_embeddings,
is_multimodal=is_multimodal,
)
def _process_image_input(self, image_input) -> tuple[torch.Tensor, ...]:
grid_thw = image_input["image_grid_thw"]