Update app.py

app.py

@@ -6,8 +6,9 @@ from transformers import AutoTokenizer, AutoModelForCausalLM
 import gradio as gr
 from fastapi import FastAPI
 from pydantic import BaseModel
+import uvicorn
 
-# Модель
+# === Модель ===
 model_id = "sberbank-ai/rugpt3medium_based_on_gpt2"
 tokenizer = AutoTokenizer.from_pretrained(model_id)
 model = AutoModelForCausalLM.from_pretrained(model_id)
@@ -24,6 +25,7 @@ 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,
@@ -33,17 +35,20 @@ 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()
     else:
         answer = output[len(prompt):].strip()
+
     return answer
 
-# Gradio интерфейс
-chat = gr.ChatInterface(fn=respond, title="Иннополис Бот")
+# === Gradio интерфейс ===
+chat_ui = gr.ChatInterface(fn=respond, title="Иннополис Бот")
 
-# API
+# === FastAPI для API ===
 app = FastAPI()
 
 class QuestionRequest(BaseModel):
@@ -53,5 +58,6 @@ class QuestionRequest(BaseModel):
 def ask(request: QuestionRequest):
     return {"answer": respond(request.question)}
 
-# Важно: экспорт для Hugging Face
-demo = gr.mount_gradio_app(app, chat, path="/")
+# === Mount Gradio UI на FastAPI ===
+# Этот объект должен быть экспортирован как `demo`
+demo = gr.mount_gradio_app(app, chat_ui, path="/")