app.py

import gradio as gr
import spaces
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM

# Load the model and tokenizer
tokenizer = AutoTokenizer.from_pretrained("TheBloke/Chronoboros-33B-GPTQ")
model = AutoModelForCausalLM.from_pretrained("TheBloke/Chronoboros-33B-GPTQ", device_map="auto")
model.eval()  # set model to evaluation mode

# Optional: Use torch.compile() if you're on PyTorch 2.0+ for further speed-up
# model = torch.compile(model)

@spaces.GPU
def respond(message, history: list[tuple[str, str]], system_message, max_tokens, temperature, top_p):
    # Build the prompt using conversation history
    prompt = f"{system_message}\n"
    for user_text, assistant_text in history:
        if user_text:
            prompt += f"User: {user_text}\n"
        if assistant_text:
            prompt += f"Assistant: {assistant_text}\n"
    prompt += f"User: {message}\nAssistant: "

    # Tokenize the prompt
    input_ids = tokenizer.encode(prompt, return_tensors="pt").to(model.device)
    
    # Generate the response with no gradients
    with torch.no_grad():
        output_ids = model.generate(
            input_ids,
            max_new_tokens=max_tokens,
            temperature=temperature,
            top_p=top_p,
            do_sample=True,
        )
    
    # Extract the new tokens
    new_tokens = output_ids[0][input_ids.shape[1]:]
    
    # Stream output in chunks (e.g., 5 tokens per chunk)
    chunk_size = 5
    for i in range(0, new_tokens.shape[0], chunk_size):
        current_response = tokenizer.decode(new_tokens[: i + chunk_size], skip_special_tokens=True)
        yield current_response

# Configure the ChatInterface with additional inputs
demo = gr.ChatInterface(
    respond,
    additional_inputs=[
        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
        gr.Slider(minimum=1, maximum=2048, 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 (nucleus sampling)",
        ),
    ],
)

if __name__ == "__main__":
    demo.launch()