app.py

import gradio as gr
from huggingface_hub import InferenceClient

"""
Copied from inference in colab notebook
"""

from transformers import AutoTokenizer , AutoModelForCausalLM , TextIteratorStreamer
import torch
from threading import Thread

# Load model and tokenizer globally to avoid reloading for every request
model_path = "Heit39/llama_lora_model_1"

# Load tokenizer
tokenizer = AutoTokenizer.from_pretrained(model_path, use_fast=True, legacy=False)

# Load the base model (e.g., LLaMA)
base_model = AutoModelForCausalLM.from_pretrained("unsloth/Llama-3.2-3B-Instruct")

# Load LoRA adapter
from peft import PeftModel
model = PeftModel.from_pretrained(base_model, model_path)


# Define the response function
# def respond(
#     message: str,
#     history: list[tuple[str, str]],
#     system_message: str,
#     max_tokens: int,
#     temperature: float,
#     top_p: float,
# ):
#     # Combine system message and history into a single prompt
#     messages = [{"role": "system", "content": system_message}]
#     for val in history:
#         if val[0]:
#             messages.append({"role": "user", "content": val[0]})
#         if val[1]:
#             messages.append({"role": "assistant", "content": val[1]})
#     messages.append({"role": "user", "content": message})
    
#     # Create a single text prompt from the messages
#     prompt = ""
#     for msg in messages:
#         if msg["role"] == "system":
#             prompt += f"[System]: {msg['content']}\n\n"
#         elif msg["role"] == "user":
#             prompt += f"[User]: {msg['content']}\n\n"
#         elif msg["role"] == "assistant":
#             prompt += f"[Assistant]: {msg['content']}\n\n"
    
#     # Tokenize the prompt
#     inputs = tokenizer(prompt, return_tensors="pt", truncation=True)
#     input_ids = inputs.input_ids.to("cpu")  # Ensure input is on the CPU

#     # Generate response
#     output_ids = model.generate(
#         input_ids,
#         max_length=input_ids.shape[1] + max_tokens,
#         temperature=temperature,
#         top_p=top_p,
#         do_sample=True,
#     )
    
#     # Decode the generated text
#     generated_text = tokenizer.decode(output_ids[0], skip_special_tokens=True)

#     # Extract the assistant's response from the generated text
#     assistant_response = generated_text[len(prompt):].strip()

#     # Yield responses incrementally (simulate streaming)
#     response = ""
#     for token in assistant_response.split():  # Split tokens by whitespace
#         response += token + " "
#         yield response.strip()

def respond(
    message: str,
    history: list[tuple[str, str]],
    system_message: str,
    max_tokens: int,
    temperature: float,
    top_p: float,
):
    # Combine system message and history into a single prompt
    messages = [{"role": "system", "content": system_message}]
    for val in history:
        if val[0]:
            messages.append({"role": "user", "content": val[0]})
        if val[1]:
            messages.append({"role": "assistant", "content": val[1]})
    messages.append({"role": "user", "content": message})
    
    # Tokenize the messages
    inputs = tokenizer.apply_chat_template(
        messages,
        tokenize = True,
        add_generation_prompt = True, # Must add for generation
        return_tensors = "pt",
    )
    # Generate tokens incrementally
    streamer = TextIteratorStreamer(tokenizer, skip_prompt=True)
    generation_kwargs = {
        "input_ids": inputs,
        "max_new_tokens": max_tokens,
        "temperature": temperature,
        "top_p": top_p,
        "do_sample": True,
        "streamer": streamer,
    }
    thread = Thread(target=model.generate, kwargs=generation_kwargs)
    thread.start()

    # Yield responses as they are generated
    response = ""
    for token in streamer:
        response += token
        yield response


"""
For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
"""
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()