Update app.py

app.py

@@ -15,8 +15,10 @@ try:
     torch.backends.cudnn.benchmark = True
 
     model_name = "HuggingFaceH4/zephyr-7b-beta"
+    # Pass token if required for private models.
     model = AutoModelForCausalLM.from_pretrained(
         model_name,
+        use_auth_token=HF_TOKEN,
         torch_dtype=torch.bfloat16,
         device_map="auto"
     )
@@ -24,7 +26,7 @@ try:
     if hasattr(torch, "compile"):
         model = torch.compile(model)
 
-    tokenizer = AutoTokenizer.from_pretrained(model_name)
+    tokenizer = AutoTokenizer.from_pretrained(model_name, use_auth_token=HF_TOKEN)
     inference_mode = "local"
 
 except ImportError:
@@ -37,10 +39,10 @@ except ImportError:
 
     model_name = "HuggingFaceH4/zephyr-7b-beta"
     tokenizer = AutoTokenizer.from_pretrained(model_name)
-    client = InferenceClient(model_name)
+    # Pass the token to the client to avoid authentication errors.
+    client = InferenceClient(model_name, token=HF_TOKEN)
     inference_mode = "client"
 
-
 # ------------------------------------------------------------------------------
 # SYSTEM PROMPT (PATIENT ROLE)
 # ------------------------------------------------------------------------------
@@ -63,7 +65,6 @@ BEHAVIOR INSTRUCTIONS:
 - Keep your responses concise, aiming for a maximum of {max_response_words} words.
 Start the conversation by expressing your current feelings or challenges from the patient's point of view."""
 
-
 # ------------------------------------------------------------------------------
 # Utility Functions
 # ------------------------------------------------------------------------------
@@ -91,7 +92,6 @@ def truncate_response(text: str, max_words: int) -> str:
         return " ".join(words[:max_words]) + "..."
     return text
 
-
 # ------------------------------------------------------------------------------
 # Response Function
 # ------------------------------------------------------------------------------
@@ -138,7 +138,6 @@ def respond(
         final_response = truncate_response(generated_response, max_response_words)
         return final_response
 
-
 # ------------------------------------------------------------------------------
 # Optional Initial Message and Gradio Interface
 # ------------------------------------------------------------------------------
@@ -164,4 +163,4 @@ demo = gr.ChatInterface(
 )
 
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch(share=True)