app.py

import spaces
from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
import gradio as gr
from threading import Thread

checkpoint = "WillHeld/soft-raccoon"
device = "cuda"
tokenizer = AutoTokenizer.from_pretrained(checkpoint)
model = AutoModelForCausalLM.from_pretrained(checkpoint).to(device)

@spaces.GPU(duration=120)
def predict(message, history, temperature, top_p):
    history.append({"role": "user", "content": message})
    input_text = tokenizer.apply_chat_template(history, tokenize=False, add_generation_prompt=True)
    inputs = tokenizer.encode(input_text, return_tensors="pt").to(device)
    
    # Create a streamer
    streamer = TextIteratorStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)
    
    # Set up generation parameters
    generation_kwargs = {
        "input_ids": inputs,
        "max_new_tokens": 1024,
        "temperature": float(temperature),
        "top_p": float(top_p),
        "do_sample": True,
        "streamer": streamer,
        "eos_token_id": 128009
    }
    
    # Run generation in a separate thread
    thread = Thread(target=model.generate, kwargs=generation_kwargs)
    thread.start()
    
    # Yield from the streamer as tokens are generated
    partial_text = ""
    for new_text in streamer:
        partial_text += new_text
        yield partial_text

with gr.Blocks() as demo:
    chatbot = gr.ChatInterface(
        predict,
        additional_inputs=[
            gr.Slider(0.1, 2.0, value=0.7, step=0.1, label="Temperature"),
            gr.Slider(0.1, 1.0, value=0.9, step=0.05, label="Top-P")
        ],
        type="messages"
    )

demo.launch()