Update utils/model.py

utils/model.py

@@ -427,7 +427,8 @@ class OwlViTForClassification(nn.Module):
 
             device_ = txt_embeds.device
             position_ids = position_ids.to(device_)
-            position_embedding = position_ids.repeat(text_embeds.size(0), 1, 1)
+            txt_embeds_size_0 = text_embeds.size(0)
+            position_embedding = position_ids.repeat(txt_embeds_size_0, 1, 1)
             text_inputs_parts["position_ids"] = position_ids
             print(f"position_embedding : {position_embedding.shape}")
             print(f"pos + emb: {(txt_embeds + position_embedding).shape}")