Update app.py

app.py

@@ -2,7 +2,7 @@ import torch
 import gradio as gr
 from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
 
-model_id = "google/flan-t5-base"
+model_id = "google/flan-t5-base"  # можно попробовать flan-t5-large
 tokenizer = AutoTokenizer.from_pretrained(model_id)
 model = AutoModelForSeq2SeqLM.from_pretrained(model_id)
 device = "cuda" if torch.cuda.is_available() else "cpu"
@@ -12,19 +12,19 @@ def respond(message, history=None):
     if history is None:
         history = []
 
-    prompt = f"Question: {message} Answer:"
+    prompt = f"Answer the following question about Innopolis University clearly and concisely.\nQuestion: {message}\nAnswer:"
     inputs = tokenizer(prompt, return_tensors="pt").to(device)
 
     with torch.no_grad():
         outputs = model.generate(
             **inputs,
-            max_new_tokens=50,
+            max_new_tokens=100,
             do_sample=False,
             eos_token_id=tokenizer.eos_token_id
         )
     answer = tokenizer.decode(outputs[0], skip_special_tokens=True)
 
-    # Убираем из ответа префикс prompt, если он остался
+    # Отрезаем если модель включила prompt в ответ
     if answer.lower().startswith(prompt.lower()):
         answer = answer[len(prompt):].strip()