init

app.py

@@ -109,11 +109,11 @@ def main():
     args = opt.parse_args()
     
     device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
-    model = EVPDepth(args=args, caption_aggregation=True)
-    model.to(device)
-    model_weight = torch.load('best_model_nyu.ckpt', map_location=device)['model']
-    model.load_state_dict(model_weight, strict=False)
-    model.eval()
+    #model = EVPDepth(args=args, caption_aggregation=True)
+    #model.to(device)
+    #model_weight = torch.load('best_model_nyu.ckpt', map_location=device)['model']
+    #model.load_state_dict(model_weight, strict=False)
+    #model.eval()
     tokenizer = CLIPTokenizer.from_pretrained("openai/clip-vit-large-patch14")
     model_refseg = EVPRefer()
     model_refseg.to(device)
@@ -121,6 +121,8 @@ def main():
     model_refseg.load_state_dict(model_weight, strict=False)
     model_refseg.eval()
     
+    del model_weight
+    print('Models uploaded successfully')
     
     title = "# EVP"
     description = """Official demo for **EVP: Enhanced Visual Perception using Inverse Multi-Attentive Feature
@@ -132,7 +134,7 @@ def main():
         gr.Markdown(title)
         gr.Markdown(description)
         with gr.Tab("Depth Prediction"):
-            create_depth_demo(model, device)
+            create_depth_demo(model_refseg, device)
         with gr.Tab("Referring Segmentation"):
             create_refseg_demo(model_refseg, tokenizer, device)
         gr.HTML('''<br><br><br><center>You can duplicate this Space to skip the queue:<a href="https://huggingface.co/spaces/MykolaL/evp?duplicate=true"><img src="https://bit.ly/3gLdBN6" alt="Duplicate Space"></a><br>