app.py

import gradio as gr
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, GenerationConfig, LlamaConfig
from peft import PeftModel  # For loading adapter files

# Path to the base model and adapter
BASE_MODEL_PATH = "unsloth/Llama-3.2-3B-Instruct"  # Replace with your base model path
ADAPTER_PATH = "Futuresony/future_ai_12_10_2024.gguf/adapter"  # Your Hugging Face repo

# Function to clean rope_scaling in model config
def clean_rope_scaling(config):
    if "rope_scaling" in config:
        valid_rope_scaling = {"type": "linear", "factor": config["rope_scaling"].get("factor", 1.0)}
        config["rope_scaling"] = valid_rope_scaling
    return config

# Load base model and tokenizer
print("Loading base model and tokenizer...")
tokenizer = AutoTokenizer.from_pretrained(BASE_MODEL_PATH)

# Load and clean the model config
config = LlamaConfig.from_pretrained(BASE_MODEL_PATH)
clean_config = clean_rope_scaling(config.to_dict())

# Load model with cleaned config
model = AutoModelForCausalLM.from_pretrained(BASE_MODEL_PATH, config=clean_config, torch_dtype=torch.float16, device_map="auto")

# Load adapter using PEFT
print("Loading adapter...")
model = PeftModel.from_pretrained(model, ADAPTER_PATH)

# Set model to evaluation mode
model.eval()

# Function to generate responses
def respond(
    message,
    history: list[tuple[str, str]],
    system_message,
    max_tokens,
    temperature,
    top_p,
):
    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})

    # Prepare input
    input_text = "\n".join([f"{msg['role']}: {msg['content']}" for msg in messages])
    inputs = tokenizer(input_text, return_tensors="pt").to(model.device)

    # Generate response
    generation_config = GenerationConfig(
        max_new_tokens=max_tokens,
        temperature=temperature,
        top_p=top_p,
        do_sample=True,
    )
    output_ids = model.generate(**inputs, generation_config=generation_config)
    response = tokenizer.decode(output_ids[0], skip_special_tokens=True)

    return response.split("assistant:")[-1].strip()

# Gradio Interface
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()