Update app.py

app.py

@@ -33,19 +33,26 @@ def respond(
     - max_tokens: the maximum number of tokens to generate in the response
     - temperature: sampling temperature
     - top_p: top-p (nucleus) sampling
-    - frequency_penalty: penalize repeated tokens in the response
+    - frequency_penalty: penalize repeated tokens in the output
     - seed: a fixed seed for reproducibility; -1 will mean 'random'
-    - model: the selected model
-    - custom_model: the custom model path
+    - model: the selected model from the featured list
+    - custom_model: a custom model specified by the user
     """
 
     print(f"Received message: {message}")
     print(f"History: {history}")
-    print(f"system message: {system_message}")
-    print(f"max tokens: {max_tokens}, Temperature: {temperature}, Top-P: {top_p}")
+    print(f"System message: {system_message}")
+    print(f"Max tokens: {max_tokens}, Temperature: {temperature}, Top-P: {top_p}")
     print(f"Frequency Penalty: {frequency_penalty}, Seed: {seed}")
-    print(f"Selected Model: {model}")
-    print(f"Custom model: {custom_model}")
+    print(f"Model: {model}, Custom Model: {custom_model}")
+
+    # Determine the model to use
+    if custom_model.strip() != "":
+        selected_model = custom_model.strip()
+    else:
+        selected_model = model
+
+    print(f"Selected model for inference: {selected_model}")
 
     # Convert seed to None if -1 (meaning random)
     if seed == -1:
@@ -61,7 +68,7 @@ def respond(
         if user_part:
             messages.append({"role": "user", "content": user_part})
             print(f"Added user message to context: {user_part}")
-        ifassistant_part:
+        if assistant_part:
             messages.append({"role": "assistant", "content": assistant_part})
             print(f"Added assistant message to context: {assistant_part}")
 
@@ -70,24 +77,25 @@ def respond(
 
     # Start with an empty string to build the response as tokens stream in
     response = ""
-    print("Sending request to OpenAI API.")
+    print(f"Sending request to OpenAI API using model: {selected_model}.")
 
-    # Make the request to the HF Inference API via openAI-like client
+    # Make the streaming request to the HF Inference API via openai-like client
     for message_chunk in client.chat.completions.create(
-        model=custom_model if custom_model.strip() != "" else model,
+        model=selected_model,
         max_tokens=max_tokens,
         stream=True,  # Stream the response
         temperature=temperature,
         top_p=top_p,
-        frequency_penalty=frequency_penalty,  # <--
-        seed=seed,  # <--
-        messages=messages
+        frequency_penalty=frequency_penalty,
+        seed=seed,
+        messages=messages,
     ):
         # Extract the token text from the response chunk
-        token_text = message_chunk.choices[0].message.content
-        print(f"Received token: {token_text}")
-        response += token_text
-        yield response
+        token_text = message_chunk.choices[0].delta.content
+        if token_text is not None:
+            print(f"Received token: {token_text}")
+            response += token_text
+            yield response
 
     print("Completed response generation.")
 
@@ -95,158 +103,108 @@ def respond(
 chatbot = gr.Chatbot(height=600)
 print("Chatbot interface created.")
 
-# Define the Gradio interface
-with gr.Blocks(theme='Nymbo/Nymbo_Theme') as demo:
-    # Tab for basic settings
-    with gr.Tab("Basic Settings"):
-        with gr.Column(elem_id="prompt-container"):
-            with gr.Row():
-                # Textbox for user to input the message
-                text_prompt = gr.Textbox(label="Prompt", placeholder="Enter a prompt here", lines=3, elem_id="prompt-text-input")
-            with gr.Row():
-                # Textbox for custom model input
-                custom_model = gr.textbox(label="Custom Model", info="HuggingFace model path (optional)", placeholder="meta-llama/Llama-3.3-70B-Instruct", lines=1, elem_id="model-search-input")
-            # Accordion for selecting the model
-            with gr.Accordion("Featured models", open=True):
-                # Textbox for searching models
-                model_search = gr.textbox(Label="Filter models", placeholder="Search for a featured model...", lines=1, elem_id="model-search-input")
-                # Radio buttons to select the desired model
-                model = gr.Radio(label="Select a model below", value="meta-llama/Llama-3.3-70B-Instruct", choices=[
-                    "meta-llama/Llama-3.3-70B-Instruct",
-                    "anthropic/claude-3",
-                    "anthropic/claude-instant-3",
-                    "anthropic/claude-2",
-                    "anthropic/claude-2",
-                    "anthropic/claude-instant-2",
-                    "anthropic/claude-1.3",
-                    "anthropic/claude-instant-1.3",
-                    "anthropic/claude-1",
-                    "anthropic/claude-instant-1",
-                    "anthropic/claude-0.3",
-                    "anthropic/claude-instant-0.3",
-                    "anthropic/claude-0.1",
-                    "anthropic/claude-instant-0.1",
-                    "anthropic/claude-v2",
-                    "anthropic/claude-instant-v2",
-                    "anthropic/claude-v1",
-                    "anthropic/claude-instant-v1",
-                    "anthropic/claude-v0.3",
-                    "anthropic/claude-instant-v0.3",
-                    "anthropic/claude-v0.1",
-                    "anthropic/claude-instant-v0.1",
-                ], interactive=True, elem_id="model-radio")
-
-                # Filtering models based on search input
-                def filter_models(search_term):
-                    filtered_models = [m for m in model.choices if search_term.lower() in m.lower()]
-                    return gr.update(choices=filtered_models)
-
-                # Update model list when search box is used
-                model_search.change(filter_models, inputs=model, outputs=model)
-
-    # Tab for advanced settings
-    with gr.Tab("Advanced Settings"):
-        with gr.Row():
-            # Text box for specifying the system message
-            system_message = gr.text box(value="", label="System message")
-        with gr.Row():
-            # Slider for setting the maximum new tokens
-            max_tokens = gr.Slider(minimum=1, maximum=4096, value=512, step=1, label="Max new tokens")
-        with gr.Row():
-            # Slider for setting the temperature
-            temperature = gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature")
-        with gr.Row():
-            #Slider for setting top-p
-            top_p = gr.Slider(minimum=0.1, maximum=1.0, value=0.9, step=0.05, label="Top-P")
-        with gr.Row():
-            #Slider for setting frequency penalty
-            frequency_penalty = gr.Slider(minimum=-2.0, maximum=2.0, value=0.0, step=0.1, label="Frequency Penalty")
-        with gr.Row():
-            #Slider for setting the seed
-            seed = gr.SLider(minimum=-1, maximum=65535, value=-1, step=1, label="Seed (-1 for random)")
+# Define featured models
+featured_models_list = [
+    "meta-llama/Llama-3.3-70B-Instruct",
+    "mistralai/Mistral-7B-v0.1",
+    "google/gemma-7b",
+]
 
-    # Tab for information
-    with gr.tab("Information"):
+# Create the Gradio ChatInterface
+with gr.Blocks(theme="Nymbo/Nymbo_Theme") as demo:
+    with gr.Tab("Chat"):
         with gr.Row():
-            # Display a sample prompt
-            gr.textbox(label="Sample prompt", value="Enter a prompt | ultra detail, ultra elaboration, ultra quality, perfect.")
-        with gr.Accordion("Featured Models (WiP)", open=False):
-            gr.html(
+            with gr.Column():
+                # Chat interface
+                gr.ChatInterface(
+                    respond,
+                    additional_inputs=[
+                        gr.Textbox(value="", label="System message"),
+                        gr.Slider(minimum=1, maximum=4096, value=512, step=1, label="Max new tokens"),
+                        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
+                        gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-P"),
+                        gr.Slider(minimum=-2.0, maximum=2.0, value=0.0, step=0.1, label="Frequency Penalty"),
+                        gr.Slider(minimum=-1, maximum=65535, value=-1, step=1, label="Seed (-1 for random)"),
+                        gr.Dropdown(label="Featured Models", choices=featured_models_list, value="meta-llama/Llama-3.3-70B-Instruct", interactive=True),
+                        gr.Textbox(value="", label="Custom Model (Optional)"),
+                    ],
+                    fill_height=True,
+                    chatbot=chatbot,
+                )
+            with gr.Column():
+                # Featured models accordion
+                with gr.Accordion("Featured Models", open=True):
+                    model_search = gr.Textbox(label="Filter Models", placeholder="Search for a featured model...", lines=1)
+                    model_radio = gr.Radio(label="Select a model below", choices=featured_models_list, value="meta-llama/Llama-3.3-70B-Instruct", interactive=True)
+
+                    def filter_models(search_term):
+                        filtered_models = [m for m in featured_models_list if search_term.lower() in m.lower()]
+                        return gr.update(choices=filtered_models)
+
+                    model_search.change(filter_models, inputs=model_search, outputs=model_radio)
+
+                # Custom model textbox
+                custom_model_textbox = gr.Textbox(label="Custom Model", placeholder="Enter a custom model path here (optional)", lines=1)
+
+    with gr.Tab("Information"):
+        with gr.Accordion("Featured Models", open=False):
+            gr.HTML(
+                """
+                <p><a href="https://huggingface.co/models?pipeline_tag=text-generation&sort=trending">See all available models</a></p>
+                <table style="width:100%; text-align:center; margin:auto;">
+                    <tr>
+                        <th>Model Name</th>
+                        <th>Notes</th>
+                    </tr>
+                    <tr>
+                        <td>meta-llama/Llama-3.3-70B-Instruct</td>
+                        <td>Powerful large language model.</td>
+                    </tr>
+                    <tr>
+                        <td>mistralai/Mistral-7B-v0.1</td>
+                        <td>A smaller, efficient model.</td>
+                    </tr>
+                    <tr>
+                        <td>google/gemma-7b</td>
+                        <td>Google's language model.</td>
+                    </tr>
+                </table>
                 """
-            <p><a href="https://huggingface.co/models?inferences=warm&pipeline_tag=text-to-text&sort=trending">View more models</a></p>
-            <table style="width:100%; text-align:center; margin:auto;">
-                <tr>
-                    <th>Model</th>
-                    <th>Description</th>
-                </tr>
-                <tr>
-                    <td>meta-llama/Llama-3.3-70B-Instruct</td>
-                    <td>High-quality, large-scale language model</td>
-                </tr>
-                <tr>
-                    <td>anthropic/claude-3</td>
-                    <td> Advanced conversational AI model</td>
-                </tr>
-                <tr>
-                    <td>anthropic/claude-instant-3</td>
-                    <td> Fast and efficient conversational AI model</td>
-                </tr>
-            </table>
-            """
             )
+
         with gr.Accordion("Parameters Overview", open=False):
-            gr.markdown(
-            """
-            ## System Message
-            - **Description**: The system message provides context and instructions to the model.
-            - **Default**: ""
-
-            ## Max New Tokens
-            - **Description**: The maximum number of tokens to generate in the response.
-            - **Default**: 512
-            - **Range**: 1 to 4096
-
-            ## Temperature
-            - **Description**: Controls the randomness of the output. Lower values make the output more deterministic, higher values make it output more varied.
- - **Default**: 0.7
- - **Range**: 0.1 to 4.0
-
-            ## Top-P
-            - **Description**: Controls the diversity of the output. Lower values make the output more focused, higher values make it more varied.
-            - **Default**: 0.7
-            - **Range**: 0.1 to 1.0
-
-            ## Frequency Penalty
-            - **Description**: Penalizes repeated tokens in the response. Higher values makes the output less repetitive.
- - **Default**: 0.0
- - **Range**: -2.0 to 2.0
-
-            ## Seed
-            - **Description**: A fixed seed for reproducibility. -1 for random.
-            - **Default**: -1
-            - **Range**: -1 to 65535
-
-            """
+            gr.Markdown(
+                """
+                ## Parameters Overview
+
+                ### System Message
+                The system message is an initial instruction or context that you provide to the chatbot. It sets the stage for the conversation and can be used to guide the chatbot's behavior or persona.
+
+                ### Max New Tokens
+                This parameter limits the length of the chatbot's response. It specifies the maximum number of tokens (words or subwords) that the chatbot can generate in a single response.
+
+                ### Temperature
+                Temperature controls the randomness of the chatbot's responses. A higher temperature (e.g., 1.0) makes the output more random and creative, while a lower temperature (e.g., 0.2) makes the output more focused and deterministic.
+
+                ### Top-P
+                Top-P, also known as nucleus sampling, is another way to control the randomness of the responses. It sets a threshold for the cumulative probability of the most likely tokens. The chatbot will only consider tokens whose cumulative probability is below this threshold.
+
+                ### Frequency Penalty
+                This parameter discourages the chatbot from repeating the same tokens or phrases too often. A higher value (e.g., 1.0) penalizes repetition more strongly, while a lower value (e.g., 0.0) has no penalty.
+
+                ### Seed
+                The seed is a number that initializes the random number generator used by the chatbot. If you set a specific seed, you will get the same response every time you run the chatbot with the same parameters. If you set the seed to -1, a random seed will be used, resulting in different responses each time.
+
+                ### Featured Models
+                You can select a featured model from the dropdown list. These models have been pre-selected for their performance and capabilities.
+
+                ### Custom Model
+                If you have a specific model that you want to use, you can enter its path in the Custom Model textbox. This allows you to use models that are not included in the featured list.
+                """
             )
-"""
-
-    # Row containing the 'Run' button to trigger the query function
-    with gr.Row():
-        text_button = gr.Button("Run", variant='primary', elem_id="gen-button")
-    # Row for displaying the generated response
-    with gr.Row():
-        response_output = gr.Textbox(label="Response Output", elem_id="response-output")
-
-    # Set up button to call the respond function
-    text_button.click(
-        respond,
-        inputs=[
-            text_prompt, model, custom_model, system_message, max_tokens, temperature, top_p, frequency_penalty, seed
-        ],
-        outputs=[response_output]
-    )
 
 print("Gradio interface initialized.")
 
 if __name__ == "__main__":
-    demo.launch(show_api=False, share=False)
+    print("Launching the demo application.")
+    demo.launch()