Update app.py

app.py

@@ -1,5 +1,6 @@
 import streamlit as st
-from transformers import AutoTokenizer, AutoModelForCausalLM
+from transformers import AutoModelForCausalLM
+from transformers.models.qwen2.tokenization_qwen2 import Qwen2Tokenizer
 import torch
 from PIL import Image
 import io
@@ -8,7 +9,7 @@ import io
 def load_model():
     model_name = "Qwen/Qwen2-VL-7B-Instruct"
     try:
-        tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
+        tokenizer = Qwen2Tokenizer.from_pretrained(model_name, trust_remote_code=True)
         model = AutoModelForCausalLM.from_pretrained(model_name, device_map="auto", trust_remote_code=True)
         return tokenizer, model
     except Exception as e:
@@ -22,7 +23,7 @@ def generate_response(prompt, image, tokenizer, model):
     try:
         if image:
             image = Image.open(image).convert('RGB')
-            inputs = tokenizer.from_pretrained(prompt, images=[image], return_tensors='pt').to(model.device)
+            inputs = tokenizer(prompt, images=[image], return_tensors='pt').to(model.device)
         else:
             inputs = tokenizer(prompt, return_tensors='pt').to(model.device)
         
@@ -40,6 +41,8 @@ tokenizer, model = load_model()
 
 if tokenizer is None or model is None:
     st.warning("Модель не загружена. Приложение может работать некорректно.")
+else:
+    st.success("Модель успешно загружена!")
 
 if "messages" not in st.session_state:
     st.session_state.messages = []
@@ -67,7 +70,8 @@ if prompt or uploaded_file:
             st.markdown(prompt)
     
     with st.chat_message("assistant"):
-        response = generate_response(prompt, uploaded_file, tokenizer, model)
+        with st.spinner("Генерация ответа..."):
+            response = generate_response(prompt, uploaded_file, tokenizer, model)
         st.markdown(response)
     
     st.session_state.messages.append({"role": "assistant", "content": response})