app.py

import os
os.system("pip3 install transformers")
os.system("pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu")
os.system("pip3 install tensorflow")
os.system("pip3 install accelerate")

import gradio as gr
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
import torch

def load_model():
    model = AutoModelForCausalLM.from_pretrained(
        "nvidia/Llama-3.1-Nemotron-Nano-8B-v1",
        device_map="auto", 
        torch_dtype=torch.float16, 
        low_cpu_mem_usage=True, 
        trust_remote_code=True
    )
    return model

model = load_model()
tokenizer = AutoTokenizer.from_pretrained(
    "nvidia/Llama-3.1-Nemotron-Nano-8B-v1",
    trust_remote_code=True
)

pipe = pipeline(
    "text-generation",
    model=model,
    tokenizer=tokenizer,
    device_map="auto"
)

def generate_response(request):
    try:
        messages = [
            {"role": "user", "content": str(request)},
        ]
        
        outputs = pipe(
            messages,
            max_new_tokens=512,
            do_sample=True,
            temperature=0.7,
            top_p=0.9,
            repetition_penalty=1.1
        )
        
        return outputs[0]["generated_text"][-1]['content']
    
    except Exception as e:
        return f"Ошибка генерации: {str(e)}"

demo = gr.Interface(
    fn=generate_response,
    inputs=gr.Textbox(label="Ваш запрос", placeholder="Введите вопрос..."),
    outputs=gr.Textbox(label="Ответ модели"),
    title="Chat with Llama-3.1-Nemotron-Nano (FP16)",
    description="Модель работает в режиме float16 (без 8-битного квантования)"
)

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