Update src/streamlit_app.py

src/streamlit_app.py

@@ -3,36 +3,35 @@ import streamlit as st
 import torch
 from transformers import AutoTokenizer, AutoModelForCausalLM
 
-# --- Dùng thư mục cache riêng để tránh PermissionError ---
+# --- Dùng thư mục cache riêng, tránh PermissionError ---
 os.environ["TRANSFORMERS_CACHE"] = "./hf_cache"
 os.environ["HF_HOME"] = "./hf_cache"
 
-st.title("Robotics Chatbot 💬")
+st.title("🤖 Fine-tuned Qwen3 Chatbot")
+
+# --- Model paths ---
+BASE_MODEL = "unsloth/Qwen3-4B-Instruct-2507"
+FINE_TUNED = "phuphan1310/Fine-tuned-model-test"
 
-MODEL_NAME = "phuphan1310/Fine-tuned-model-test"  # model Qwen3 4B fine-tuned
 device = "cuda" if torch.cuda.is_available() else "cpu"
 
 @st.cache_resource(show_spinner=True)
 def load_model():
-    base_model = "unsloth/Qwen3-4B-Instruct-2507"
-    fine_tuned_model = "phuphan1310/Fine-tuned-model-test"
-
+    # ⚠️ Dùng tokenizer từ model gốc (Unsloth) vì tokenizer fine-tuned lỗi format
     tokenizer = AutoTokenizer.from_pretrained(
-        base_model,
+        BASE_MODEL,
         trust_remote_code=True
     )
     model = AutoModelForCausalLM.from_pretrained(
-        fine_tuned_model,
+        FINE_TUNED,
+        trust_remote_code=True,
         torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32,
-        device_map="auto",
-        trust_remote_code=True
+        device_map="auto"
     )
     return tokenizer, model
 
-
 tokenizer, model = load_model()
 
-# --- Hàm tạo response ---
 def generate_response(prompt):
     inputs = tokenizer(prompt, return_tensors="pt").to(device)
     outputs = model.generate(
@@ -44,7 +43,6 @@ def generate_response(prompt):
     )
     return tokenizer.decode(outputs[0], skip_special_tokens=True)
 
-# --- Lưu lịch sử chat ---
 if "messages" not in st.session_state:
     st.session_state.messages = []
 
@@ -54,7 +52,6 @@ if user_input:
     response = generate_response(user_input)
     st.session_state.messages.append({"role": "assistant", "content": response})
 
-# --- Hiển thị lịch sử chat ---
 for msg in st.session_state.messages:
     if msg["role"] == "user":
         st.markdown(f"**You:** {msg['content']}")