app.py

import gradio as gr
from transformers import AutoModelForCausalLM, AutoTokenizer

# Загружаем модель и токенизатор
model_name = "nvidia/Hymba-1.5B-Instruct"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)

# Функция генерации текста
def generate_text(prompt):
    inputs = tokenizer(prompt, return_tensors="pt")
    outputs = model.generate(**inputs, max_length=100)
    return tokenizer.decode(outputs[0], skip_special_tokens=True)

# Интерфейс Gradio
interface = gr.Interface(
    fn=generate_text,
    inputs=gr.Textbox(label="Введите запрос"),
    outputs=gr.Textbox(label="Ответ")
)

interface.launch()