app.py

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

MODEL_NAME = "S1mp1eXXX/Nimi-1b-thinking"

# Load once at startup (important)
tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)

model = AutoModelForCausalLM.from_pretrained(
    MODEL_NAME,
    torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32,
    device_map="auto"
)

def respond(message, history, system_message, max_tokens, temperature, top_p):

    messages = system_message + "\n"

    for h in history:
        messages += f"{h['role']}: {h['content']}\n"

    messages += f"user: {message}\nassistant:"

    inputs = tokenizer(messages, return_tensors="pt").to(model.device)

    streamer = TextIteratorStreamer(
        tokenizer,
        skip_prompt=True,
        skip_special_tokens=True
    )

    generation_kwargs = dict(
        **inputs,
        max_new_tokens=max_tokens,
        temperature=temperature,
        top_p=top_p,
        streamer=streamer
    )

    thread = Thread(target=model.generate, kwargs=generation_kwargs)
    thread.start()

    partial_output = ""
    for new_token in streamer:
        partial_output += new_token
        yield partial_output


chatbot = gr.ChatInterface(
    respond,
    type="messages",
    additional_inputs=[
        gr.Textbox(value="You are a helpful assistant.", label="System message"),
        gr.Slider(1, 2048, value=512, step=1, label="Max new tokens"),
        gr.Slider(0.1, 4.0, value=0.7, step=0.1, label="Temperature"),
        gr.Slider(0.1, 1.0, value=0.95, step=0.05, label="Top-p"),
    ],
)

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