Api issue fix

app.py

@@ -5,40 +5,37 @@ import torch
 from transformers import AutoModelForCausalLM, AutoTokenizer
 
 # Load the tokenizer and model
-# Using a specific revision to ensure compatibility
+print("Loading DialoGPT-medium model...")
 tokenizer = AutoTokenizer.from_pretrained("microsoft/DialoGPT-medium")
 model = AutoModelForCausalLM.from_pretrained("microsoft/DialoGPT-medium")
+print("Model loaded successfully!")
 
-# Define the prediction function
-def predict(message, history):
+# Define the chat function for the modern ChatInterface
+def chat_fn(message, history):
     try:
-        # Validate inputs
+        # Validate input
         if not message or not message.strip():
-            return "", history
+            return "Please enter a message."
         
-        if history is None:
-            history = []
-        
-        # 'history' is a list of lists, where each inner list has a user and a bot message.
-        # We need to format it for DialoGPT.
+        # Format history for DialoGPT
+        # History comes as a list of [user_msg, bot_msg] pairs
         history_transformer_format = []
         for exchange in history:
-            if isinstance(exchange, list) and len(exchange) >= 2:
+            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))
 
-        # Join the history and the new message, separated by the EOS token
+        # Create the input text
         history_string = "".join(history_transformer_format)
         input_text = history_string + str(message) + tokenizer.eos_token
 
         # Tokenize the input
         new_user_input_ids = tokenizer.encode(input_text, return_tensors='pt')
 
-        # Generate a response
-        # The max_length is set to 1250 to allow for a decent conversation history.
+        # Generate a response with memory management
         with torch.no_grad():
             bot_output_ids = model.generate(
                 new_user_input_ids,
@@ -48,41 +45,54 @@ def predict(message, history):
                 do_sample=True,
                 top_k=100,
                 top_p=0.7,
-                temperature=0.8
+                temperature=0.8,
+                early_stopping=True
             )
 
-        # Decode the response, skipping the input part
-        response = tokenizer.decode(bot_output_ids[:, new_user_input_ids.shape[-1]:][0], skip_special_tokens=True)
+        # Decode the response
+        response = tokenizer.decode(
+            bot_output_ids[:, new_user_input_ids.shape[-1]:][0], 
+            skip_special_tokens=True
+        ).strip()
         
-        # Clean up response
-        response = response.strip()
+        # Fallback for empty responses
         if not response:
-            response = "I'm not sure how to respond to that. Could you try rephrasing?"
+            response = "I'm not sure how to respond to that. Could you try rephrasing your question?"
         
-        # Return an empty string to clear the textbox and the updated history
-        return "", history + [[message, response]]
+        return response
         
     except Exception as e:
-        print(f"Error in predict function: {e}")
-        error_response = "Sorry, I encountered an error. Please try again."
-        return "", history + [[message, error_response]]
-
-# Build the Gradio interface
-with gr.Blocks() as demo:
-    gr.Markdown("## DialoGPT-medium Chatbot")
-    gr.Markdown("This chatbot uses the microsoft/DialoGPT-medium model. Start typing to chat!")
-    
-    chatbot = gr.Chatbot(value=[], label="DialoGPT Conversation")
-    textbox = gr.Textbox(placeholder="Type your message here and press Enter", label="Message")
+        print(f"Error in chat function: {e}")
+        return "Sorry, I encountered an error processing your message. Please try again."
 
-    # When the user submits the textbox, call the 'predict' function
-    textbox.submit(
-        predict,
-        inputs=[textbox, chatbot],
-        outputs=[textbox, chatbot]
+# Create the Gradio 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!",
+    examples=[
+        "Hello, how are you?",
+        "What's your favorite movie?",
+        "Tell me a joke",
+        "What do you think about artificial intelligence?"
+    ],
+    cache_examples=False,
+    retry_btn="🔄 Retry",
+    undo_btn="↶ Undo",
+    clear_btn="🗑️ Clear",
+    submit_btn="Send",
+    textbox=gr.Textbox(
+        placeholder="Type your message here...",
+        container=False,
+        scale=7
     )
+)
 
-# Enable the queue for better handling of multiple users and to enable API usage
-demo.queue()
 # Launch the app
-demo.launch()
+if __name__ == "__main__":
+    demo.queue(max_size=20)  # Enable queue for better concurrent handling
+    demo.launch(
+        server_name="0.0.0.0",
+        server_port=7860,
+        share=False
+    )

requirements.txt

@@ -1,3 +1,3 @@
-torch
-transformers
-gradio>=3.50.0,<4.0.0
+torch>=1.9.0
+transformers>=4.21.0
+gradio>=4.0.0