app.py

import gradio as gr
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel

BASE_MODEL = "deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B"
LORA_PATH = "./"

# Tokenizer
tokenizer = AutoTokenizer.from_pretrained(BASE_MODEL)
tokenizer.pad_token = tokenizer.eos_token

# Base model (CPU)
model = AutoModelForCausalLM.from_pretrained(
    BASE_MODEL,
    torch_dtype=torch.float32,
    device_map={"": "cpu"},
    low_cpu_mem_usage=True
)

# Load LoRA
model = PeftModel.from_pretrained(model, LORA_PATH)
model.eval()


def chat(user_prompt, max_tokens, temperature):
    prompt = f"""
You are a lab assistant.
Answer in **Markdown** format.
Use headings, bullet points, and code blocks when appropriate.

Question:
{user_prompt}

Answer:
"""

    inputs = tokenizer(prompt, return_tensors="pt")

    with torch.no_grad():
        output = model.generate(
            **inputs,
            max_new_tokens=int(max_tokens),
            do_sample=False,          # CPU için hızlı
            eos_token_id=tokenizer.eos_token_id
        )

    generated = output[0][inputs["input_ids"].shape[-1]:]
    return tokenizer.decode(generated, skip_special_tokens=True)


# Gradio UI
demo = gr.Interface(
    fn=chat,
    inputs=[
        gr.Textbox(lines=5, label="Prompt"),
        gr.Slider(32, 512, value=256, step=32, label="Max tokens"),
        gr.Slider(0.1, 1.5, value=0.7, step=0.1, label="Temperature"),
    ],
    outputs=gr.Markdown(label="Answer"),
    title="DeepSeek Lab Assistant (LoRA)",
)

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