Update app.py

app.py

@@ -11,24 +11,46 @@ import cv2
 import numpy as np
 import gradio as gr
 import spaces
+from huggingface_hub import login
+import os
+
+# Add login function at the start
+def init_huggingface_auth():
+    # Get token from environment variable or set it directly
+    token = os.getenv("HUGGINGFACE_TOKEN")
+    if token:
+        login(token=token)
+    else:
+        print("Warning: HUGGINGFACE_TOKEN not found in environment variables")
 
 # Load both models and their processors/tokenizers
 def load_models():
+    # Initialize HF auth before loading models
+    init_huggingface_auth()
+    
     # Vision model
     vision_model = Qwen2VLForConditionalGeneration.from_pretrained(
         "Qwen/Qwen2-VL-2B-Instruct",
         torch_dtype=torch.float16,
-        device_map="auto"
+        device_map="auto",
+        use_auth_token=True  # Add auth token usage
+    )
+    vision_processor = AutoProcessor.from_pretrained(
+        "Qwen/Qwen2-VL-2B-Instruct",
+        use_auth_token=True  # Add auth token usage
     )
-    vision_processor = AutoProcessor.from_pretrained("Qwen/Qwen2-VL-2B-Instruct")
     
-    # Code model
+    # Code model  
     code_model = AutoModelForCausalLM.from_pretrained(
         "Qwen/Qwen2.5-Coder-1.5B-Instruct",
         torch_dtype=torch.float16,
-        device_map="auto"
+        device_map="auto",
+        use_auth_token=True  # Add auth token usage
+    )
+    code_tokenizer = AutoTokenizer.from_pretrained(
+        "Qwen/Qwen2.5-Coder-1.5B-Instruct",
+        use_auth_token=True  # Add auth token usage
     )
-    code_tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2.5-Coder-1.5B-Instruct")
     
     return vision_model, vision_processor, code_model, code_tokenizer