api error 3

app.py

@@ -4,33 +4,44 @@ import gradio as gr
 import torch
 from transformers import AutoModelForCausalLM, AutoTokenizer
 
-# Load the tokenizer and model
 print("Loading DialoGPT-medium model...")
+
+# Load the tokenizer and model
 tokenizer = AutoTokenizer.from_pretrained("microsoft/DialoGPT-medium")
 model = AutoModelForCausalLM.from_pretrained("microsoft/DialoGPT-medium")
+
 print("Model loaded successfully!")
 
-# Define the chat function for the modern ChatInterface
-def chat_fn(message, history):
+# Define the prediction function that works with the modern format
+def predict(message, history):
     try:
         # Validate input
         if not message or not message.strip():
             return "Please enter a message."
         
-        # Format history for DialoGPT
-        # History comes as a list of [user_msg, bot_msg] pairs
-        history_transformer_format = []
-        for exchange in history:
-            if len(exchange) >= 2:
-                user_msg, bot_msg = exchange[0], exchange[1]
-                if user_msg:
-                    history_transformer_format.append(str(user_msg))
-                if bot_msg:
-                    history_transformer_format.append(str(bot_msg))
-
-        # Create the input text
-        history_string = "".join(history_transformer_format)
-        input_text = history_string + str(message) + tokenizer.eos_token
+        # Format history for DialoGPT - handle both old and new formats
+        history_transformer_format = ""
+        
+        # Handle the new 'messages' format (list of dicts)
+        if history and isinstance(history[0], dict):
+            for turn in history:
+                if turn.get("role") == "user":
+                    history_transformer_format += turn["content"] + tokenizer.eos_token
+                elif turn.get("role") == "assistant":
+                    history_transformer_format += turn["content"] + tokenizer.eos_token
+        
+        # Handle the old 'tuples' format (list of lists)
+        elif history and isinstance(history[0], list):
+            for exchange in history:
+                if len(exchange) >= 2:
+                    user_msg, bot_msg = exchange[0], exchange[1]
+                    if user_msg:
+                        history_transformer_format += str(user_msg) + tokenizer.eos_token
+                    if bot_msg:
+                        history_transformer_format += str(bot_msg) + tokenizer.eos_token
+        
+        # Add the current message
+        input_text = history_transformer_format + str(message) + tokenizer.eos_token
 
         # Tokenize the input
         new_user_input_ids = tokenizer.encode(input_text, return_tensors='pt')
@@ -39,14 +50,13 @@ def chat_fn(message, history):
         with torch.no_grad():
             bot_output_ids = model.generate(
                 new_user_input_ids,
-                max_length=1250,
+                max_length=1000,  # Reduced for better performance
                 pad_token_id=tokenizer.eos_token_id,
                 no_repeat_ngram_size=3,
                 do_sample=True,
-                top_k=100,
+                top_k=50,
                 top_p=0.7,
-                temperature=0.8,
-                early_stopping=True
+                temperature=0.8
             )
 
         # Decode the response
@@ -55,35 +65,38 @@ def chat_fn(message, history):
             skip_special_tokens=True
         ).strip()
         
-        # Fallback for empty responses
+        # Clean up and validate response
         if not response:
-            response = "I'm not sure how to respond to that. Could you try rephrasing your question?"
+            response = "I'm sorry, I couldn't generate a response. Could you try rephrasing your question?"
+        
+        # Limit response length to prevent protocol errors
+        if len(response) > 500:
+            response = response[:500] + "..."
         
         return response
         
     except Exception as e:
-        print(f"Error in chat function: {e}")
-        return "Sorry, I encountered an error processing your message. Please try again."
+        print(f"Error in predict function: {str(e)}")
+        return "Sorry, I encountered an error. Please try again with a different message."
 
-# Create the Gradio ChatInterface
+# Create a simple ChatInterface
 demo = gr.ChatInterface(
-    fn=chat_fn,
-    title="🤖 DialoGPT-medium Chatbot",
-    description="Chat with Microsoft's DialoGPT-medium model. This conversational AI can engage in natural dialogue!",
+    fn=predict,
+    title="DialoGPT-medium Chatbot",
+    description="Chat with Microsoft's DialoGPT-medium model!",
     examples=[
-        "Hello, how are you?",
-        "What's your favorite movie?",
+        "Hello!",
+        "How are you?",
         "Tell me a joke",
-        "What do you think about artificial intelligence?"
+        "What's the weather like?"
     ],
     cache_examples=False
 )
 
-# Launch the app
+# Launch the app with public sharing enabled
 if __name__ == "__main__":
-    demo.queue(max_size=20)  # Enable queue for better concurrent handling
     demo.launch(
+        share=True,  # This creates the public link
         server_name="0.0.0.0",
-        server_port=7860,
-        share=False
+        server_port=7860
     )