Update utils/model.py

utils/model.py

@@ -428,10 +428,10 @@ class OwlViTForClassification(nn.Module):
             device_ = txt_embeds.device
             position_ids = position_ids.to(device_)
             txt_embeds_size_0 = text_embeds.size(0)
-            position_embedding = position_ids.repeat(txt_embeds_size_0, 1, 1)
+            position_embedding = position_ids.cpu().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}")
+            print(f"pos + emb: {(txt_embeds.cpu() + position_embedding).shape}")
             text_embeds_parts = self.owlvit.text_model.get_text_features(**text_inputs_parts)
             
             # # Embed images and text queries