Update app.py

app.py

@@ -191,14 +191,53 @@ AutoModelForCausalLM.register(SmolLM2Config, SmolLM2ForCausalLM)
 # Load model and tokenizer
 model_id = "jatingocodeo/SmolLM2"
 
+def check_huggingface_access():
+    try:
+        from huggingface_hub import HfApi
+        api = HfApi()
+        
+        # Check if token exists
+        try:
+            token = os.getenv("HF_TOKEN")
+            if not token:
+                print("× Warning: HF_TOKEN environment variable not found")
+                print("  Please ensure you have set your HuggingFace token")
+                return False
+        except Exception as e:
+            print(f"× Error checking HF_TOKEN: {str(e)}")
+            return False
+        
+        # Check repository access
+        try:
+            print(f"Checking access to repository: {model_id}")
+            repo_info = api.repo_info(model_id, token=token)
+            print("✓ Repository access confirmed")
+            print(f"Repository info: {repo_info}")
+            return True
+        except Exception as e:
+            print(f"× Error accessing repository: {str(e)}")
+            return False
+    except Exception as e:
+        print(f"× Error in HuggingFace access check: {str(e)}")
+        return False
+
 def load_model():
     try:
         print("\n=== Starting model loading process ===")
         print(f"Model ID: {model_id}")
         
+        # Check HuggingFace access first
+        print("\nChecking HuggingFace access...")
+        if not check_huggingface_access():
+            raise Exception("Unable to access HuggingFace repository. Please check your token and repository permissions.")
+        
         print("\n1. Loading tokenizer...")
         try:
-            tokenizer = AutoTokenizer.from_pretrained(model_id)
+            tokenizer = AutoTokenizer.from_pretrained(
+                model_id,
+                use_auth_token=os.getenv("HF_TOKEN"),
+                trust_remote_code=True
+            )
             print("✓ Tokenizer loaded successfully")
             print(f"Tokenizer type: {type(tokenizer)}")
             print(f"Vocabulary size: {len(tokenizer)}")
@@ -235,25 +274,27 @@ def load_model():
         
         print("\n4. Loading model from Hub...")
         try:
+            # First try to list files in the repository
+            from huggingface_hub import list_repo_files
+            try:
+                files = list_repo_files(model_id, token=os.getenv("HF_TOKEN"))
+                print(f"Files in repository: {files}")
+            except Exception as hub_e:
+                print(f"Warning: Could not list repository files: {str(hub_e)}")
+            
             model = AutoModelForCausalLM.from_pretrained(
                 model_id,
                 config=config,
                 torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32,
                 trust_remote_code=True,
+                use_auth_token=os.getenv("HF_TOKEN"),
                 low_cpu_mem_usage=True,
-                local_files_only=False  # Force download from Hub
+                local_files_only=False
             )
             print("✓ Model loaded successfully")
             print(f"Model type: {type(model)}")
         except Exception as e:
             print(f"× Error loading model: {str(e)}")
-            print("Attempting to print model files in Hub repo...")
-            from huggingface_hub import list_repo_files
-            try:
-                files = list_repo_files(model_id)
-                print(f"Files in repo: {files}")
-            except Exception as hub_e:
-                print(f"Error listing repo files: {str(hub_e)}")
             raise
         
         print("\n5. Moving model to device...")
@@ -293,6 +334,8 @@ def load_model():
         print(f"CUDA available: {torch.cuda.is_available()}")
         if torch.cuda.is_available():
             print(f"CUDA version: {torch.version.cuda}")
+        print("\nEnvironment variables:")
+        print(f"HF_TOKEN set: {'HF_TOKEN' in os.environ}")
         raise
 
 def generate_text(prompt, max_length=100, temperature=0.7, top_k=50):