fix old version compatibility

huggingface · Sep 27, 2024 · 96d8de4 · 96d8de4
1 parent 2fb59ce
commit 96d8de4
Showing 1 changed file with 3 additions and 0 deletions.
diff --git a/optimum/intel/openvino/modeling_visual_language.py b/optimum/intel/openvino/modeling_visual_language.py
@@ -761,6 +761,7 @@ def merge_vision_text_embeddings(
         inputs_embeds = torch.from_numpy(inputs_embeds) if isinstance(inputs_embeds, np.ndarray) else inputs_embeds
         if legacy_processing is None:
             legacy_processing = (
+                hasattr(self.config, "image_seq_len") and
                 (input_ids == self.config.image_token_index).sum(1).max() < self.config.image_seq_length
             ) or (input_ids.shape[-1] == 1)
 
@@ -842,6 +843,7 @@ def get_multimodal_embeddings(
         self, input_ids, pixel_values=None, attention_mask=None, position_ids=None, past_key_values=None, **kwargs
     ):
         legacy_processing = (
+            hasattr(self.config, "image_seq_len") and
             (input_ids == self.config.image_token_index).sum(1).max() < self.config.image_seq_length
         ) or (input_ids.shape[-1] == 1 and pixel_values is not None)
         inputs_embeds, attention_mask, position_ids = super().get_multimodal_embeddings(
@@ -961,6 +963,7 @@ def get_multimodal_embeddings(
         inputs_embeds = self.get_text_embeddings(input_ids, **kwargs)
 
         legacy_processing = (
+            hasattr(self.config, "image_seq_len") and
             (input_ids == self.config.image_token_index).sum(1).max() < self.config.image_seq_length
         ) or (input_ids.shape[-1] == 1 and pixel_values is not None)
         if pixel_values is not None and pixel_values.size(0) > 0: