Update app.py

app.py

@@ -17,50 +17,34 @@ model_name_to_path = {
 }
 
 # Load Hugging Face token
-hf_token = os.environ["HUGGINGFACEHUB_API_TOKEN"]
+hf_token = os.environ.get("HUGGINGFACEHUB_API_TOKEN", "default_token")  # Use default to avoid errors
 
 # Preload models and tokenizers
 loaded_models = defaultdict(dict)
 
 for name, path in model_name_to_path.items():
-    loaded_models[name]["tokenizer"] = AutoTokenizer.from_pretrained(path, token=hf_token)
-    loaded_models[name]["model"] = AutoModelForCausalLM.from_pretrained(path, token=hf_token)
-    loaded_models[name]["model"].eval()
+    try:
+        loaded_models[name]["tokenizer"] = AutoTokenizer.from_pretrained(path, token=hf_token)
+        loaded_models[name]["model"] = AutoModelForCausalLM.from_pretrained(path, token=hf_token)
+        loaded_models[name]["model"].eval()
+    except Exception as e:
+        print(f"Error loading {name}: {str(e)}")
 
 def respond(message, history, model_name, max_tokens, temperature, top_p):
-    """
-    Generate a response from the selected model, streaming the output and updating chat history.
-    
-    Args:
-        message (str): User's input message.
-        history (list): Current chat history as list of (user_msg, bot_msg) tuples.
-        model_name (str): Selected model name.
-        max_tokens (int): Maximum number of tokens to generate.
-        temperature (float): Sampling temperature.
-        top_p (float): Top-p sampling parameter.
-    
-    Yields:
-        list: Updated chat history with the user's message and streaming bot response.
-    """
-    # Append user's message to history with an empty bot response
     history = history + [(message, "")]
-    yield history  # Display user's message immediately
+    yield history
 
-    # Select tokenizer and model
     tokenizer = loaded_models[model_name]["tokenizer"]
     model = loaded_models[model_name]["model"]
 
-    # Tokenize input
     inputs = tokenizer(message, return_tensors="pt")
 
-    # Set up streaming
     streamer = TextIteratorStreamer(
         tokenizer,
         skip_prompt=False,
         skip_special_tokens=True,
     )
 
-    # Configure generation parameters
     generate_kwargs = dict(
         **inputs,
         streamer=streamer,
@@ -71,53 +55,32 @@ def respond(message, history, model_name, max_tokens, temperature, top_p):
         eos_token_id=tokenizer.eos_token_id,
     )
 
-    # Start generation in a background thread
     thread = threading.Thread(target=model.generate, kwargs=generate_kwargs)
     thread.start()
 
-    # Stream the response with model name prefix
-    accumulated = f"**{model_name}:** "
+    accumulated = ""  # Removed model name prefix
     for new_text in streamer:
         accumulated += new_text
         history[-1] = (message, accumulated)
         yield history
 
 def submit(message, history, model_name, max_tokens, temperature, top_p):
-    """
-    Handle form submission by calling respond and clearing the input box.
-    
-    Args:
-        message (str): User's input message.
-        history (list): Current chat history.
-        model_name (str): Selected model name.
-        max_tokens (int): Max tokens parameter.
-        temperature (float): Temperature parameter.
-        top_p (float): Top-p parameter.
-    
-    Yields:
-        tuple: (updated chat history, cleared user input)
-    """
     for updated_history in respond(message, history, model_name, max_tokens, temperature, top_p):
         yield updated_history, ""
 
-# Create the Gradio interface with Blocks
 with gr.Blocks(css=".gr-button {margin: 5px; width: 100%;} .gr-column {padding: 10px;}") as demo:
-    # Title and description
     gr.Markdown("# LeCarnet")
     gr.Markdown("Select a model on the right and type a message to chat.")
 
-    # Two-column layout with specific widths
     with gr.Row():
-        # Left column: Chat interface (80% width)
         with gr.Column(scale=4):
             chatbot = gr.Chatbot(
-                avatar_images=(None, "media/le-carnet.png"),  # User avatar: None, Bot avatar: Logo
+                avatar_images=(None, "https://raw.githubusercontent.com/maxlsb/le-carnet/main/media/le-carnet.png"),   # Using URL for reliability
                 label="Chat",
-                height=600,  # Increase chat height for larger display
+                height=600,
             )
             user_input = gr.Textbox(placeholder="Type your message here...", label="Message")
             submit_btn = gr.Button("Send")
-            # Example prompts
             examples = gr.Examples(
                 examples=[
                     ["Il était une fois un petit garçon qui vivait dans un village paisible."],
@@ -127,25 +90,29 @@ with gr.Blocks(css=".gr-button {margin: 5px; width: 100%;} .gr-column {padding:
                 inputs=user_input,
             )
 
-        # Right column: Model selection and parameters (20% width)
         with gr.Column(scale=1, min_width=200):
-            # Dropdown for model selection
             model_dropdown = gr.Dropdown(
                 choices=["LeCarnet-3M", "LeCarnet-8M", "LeCarnet-21M"],
                 value="LeCarnet-8M",
                 label="Select Model"
             )
-            # Sliders for parameters
             max_tokens = gr.Slider(1, 512, value=512, step=1, label="Max New Tokens")
             temperature = gr.Slider(0.1, 2.0, value=0.7, step=0.1, label="Temperature")
             top_p = gr.Slider(0.1, 1.0, value=0.9, step=0.05, label="Top-p")
 
-    # Event handling for submit button
+    # Submit button click
     submit_btn.click(
         fn=submit,
         inputs=[user_input, chatbot, model_dropdown, max_tokens, temperature, top_p],
         outputs=[chatbot, user_input],
     )
+    
+    # Enter key press
+    user_input.submit(
+        fn=submit,
+        inputs=[user_input, chatbot, model_dropdown, max_tokens, temperature, top_p],
+        outputs=[chatbot, user_input],
+    )
 
 if __name__ == "__main__":
     demo.queue(default_concurrency_limit=10, max_size=10).launch(ssr_mode=False, max_threads=10)