Update utils/model.py

utils/model.py

@@ -425,6 +425,8 @@ class OwlViTForClassification(nn.Module):
             print(f"position_embedding: {self.owlvit.text_model.embeddings.position_embedding(position_ids).shape}")
             print(f"text_embeds: {txt_embeds.shape}")
             print(f"pos + emb: {(txt_embeds + position_ids).shape}")
+            text_inputs_parts["position_ids"] = position_ids
+            position_ids = position_ids.repeat(1, 1, txt_embeds.size(-1) // position_ids.size(-1))
             text_embeds_parts = self.owlvit.text_model.get_text_features(**text_inputs_parts)
             
             # # Embed images and text queries