Update app.py

app.py

@@ -2,12 +2,12 @@ import os
 os.environ["HF_HUB_ENABLE_HF_TRANSFER"] = "0"
 
 import torch
-from fastapi import FastAPI
-from pydantic import BaseModel
 from transformers import AutoTokenizer, AutoModelForCausalLM
 import gradio as gr
+from fastapi import FastAPI
+from pydantic import BaseModel
 
-# --- Модель ---
+# Модель
 model_id = "sberbank-ai/rugpt3medium_based_on_gpt2"
 tokenizer = AutoTokenizer.from_pretrained(model_id)
 model = AutoModelForCausalLM.from_pretrained(model_id)
@@ -24,7 +24,6 @@ context = (
 def respond(message, history=None):
     prompt = f"Прочитай текст и ответь на вопрос:\n\n{context}\n\nВопрос: {message}\nОтвет:"
     input_ids = tokenizer(prompt, return_tensors="pt").input_ids.to(device)
-
     with torch.no_grad():
         output_ids = model.generate(
             input_ids,
@@ -34,7 +33,6 @@ def respond(message, history=None):
             do_sample=True,
             pad_token_id=tokenizer.eos_token_id
         )
-
     output = tokenizer.decode(output_ids[0], skip_special_tokens=True)
     if "Ответ:" in output:
         answer = output.split("Ответ:")[-1].strip()
@@ -42,10 +40,10 @@ def respond(message, history=None):
         answer = output[len(prompt):].strip()
     return answer
 
-# --- Gradio UI ---
+# Gradio интерфейс
 chat = gr.ChatInterface(fn=respond, title="Иннополис Бот")
 
-# --- FastAPI для API доступа ---
+# API
 app = FastAPI()
 
 class QuestionRequest(BaseModel):
@@ -53,10 +51,7 @@ class QuestionRequest(BaseModel):
 
 @app.post("/ask")
 def ask(request: QuestionRequest):
-    answer = respond(request.question)
-    return {"answer": answer}
-
-# --- Подключаем FastAPI к Gradio ---
-gr.mount_gradio_app(app, chat, path="/")
+    return {"answer": respond(request.question)}
 
-# --- Не нужно писать iface.launch()! Hugging Face сам запустит Uvicorn ---
+# Важно: экспорт для Hugging Face
+demo = gr.mount_gradio_app(app, chat, path="/")