Update app.py

change model to v3

app.py

@@ -13,7 +13,7 @@ from transformers import (
 st.set_page_config(page_title="😶‍🌫️ FuseChat Model")
 
 root_path = "FuseAI"
-model_name = "FuseAI/FuseChat-7B-v2.0"
+model_name = "FuseAI/FuseChat-Qwen-2.5-7B-Instruct"
 
 @st.cache_resource
 def load_model(model_name):
@@ -41,20 +41,21 @@ def load_model(model_name):
 
 
 with st.sidebar:
-    st.title('😶‍🌫️ FuseChat-v2.0')
+    st.title('😶‍🌫️ FuseChat-3.0')
     st.write('This chatbot is created using FuseChat, a model developed by FuseAI')
-    temperature = st.sidebar.slider('temperature', min_value=0.01, max_value=5.0, value=0.1, step=0.01)
-    top_p = st.sidebar.slider('top_p', min_value=0.01, max_value=1.0, value=0.9, step=0.01)
-    top_k = st.sidebar.slider('top_k', min_value=1, max_value=1000, value=50, step=1)
-    repetition_penalty = st.sidebar.slider('repetition penalty', min_value=1., max_value=2., value=1.2, step=0.05)
-    max_length = st.sidebar.slider('max new tokens', min_value=32, max_value=2000, value=512, step=8)
+    temperature = st.sidebar.slider('temperature', min_value=0.01, max_value=1.0, value=0.7, step=0.01)
+    top_p = st.sidebar.slider('top_p', min_value=0.1, max_value=1.0, value=0.8, step=0.05)
+    top_k = st.sidebar.slider('top_k', min_value=1, max_value=1000, value=20, step=1)
+    repetition_penalty = st.sidebar.slider('repetition penalty', min_value=1.0, max_value=2.0, value=1.05, step=0.05)
+    max_length = st.sidebar.slider('max_length', min_value=32, max_value=4096, value=2048, step=8)
 
 with st.spinner('loading model..'):
     model, tokenizer = load_model(model_name)
 
 # Store LLM generated responses
 if "messages" not in st.session_state.keys():
-    st.session_state.messages = [{"role": "assistant", "content": "How may I assist you today?"}]
+    # st.session_state.messages = [{"role": "assistant", "content": "How may I assist you today?"}]
+    st.session_state.messages = []
 
 # Display or clear chat messages
 for message in st.session_state.messages:
@@ -64,7 +65,7 @@ for message in st.session_state.messages:
 def set_query(query):
     st.session_state.messages.append({"role": "user", "content": query})
 # Create a list of candidate questions
-candidate_questions = ["Can you tell me a joke?", "Write a quicksort code in Python.", "Write a poem about love in Shakespearean tone."]
+candidate_questions = ["Is boiling water (100°C) an obtuse angle (larger than 90 degrees)?", "Write a quicksort code in Python.", "笼子里有好几只鸡和兔子。笼子里有72个头，200只腿。里面有多少只鸡和兔子"]
 # Display the chat interface with a list of clickable question buttons
 for question in candidate_questions:
     st.sidebar.button(label=question, on_click=set_query, args=[question])
@@ -75,15 +76,15 @@ st.sidebar.button('Clear Chat History', on_click=clear_chat_history)
 
 
 def generate_fusechat_response():
-    # string_dialogue = "You are a helpful and harmless assistant."
-    string_dialogue = ""
+    conversations=[]
+    conversations.append({"role": "system", "content": "You are FuseChat-3.0, created by Sun Yat-sen University. You are a helpful assistant."})
     for dict_message in st.session_state.messages:
         if dict_message["role"] == "user":
-            string_dialogue += "GPT4 Correct User: " + dict_message["content"] + "<|end_of_turn|>"
+            conversations.append({"role": "user", "content": dict_message["content"]})
         else:
-            string_dialogue += "GPT4 Correct Assistant: " + dict_message["content"] + "<|end_of_turn|>"
-
-    input_ids = tokenizer(f"{string_dialogue}GPT4 Correct Assistant: ", return_tensors="pt").input_ids
+            conversations.append({"role": "assistant", "content": dict_message["content"]})
+    string_dialogue = tokenizer.apply_chat_template(conversations, tokenize=False, add_generation_prompt=True)
+    input_ids = tokenizer(string_dialogue, return_tensors="pt").input_ids.to('cuda')
     streamer = TextIteratorStreamer(tokenizer, timeout=10.0, skip_prompt=True, skip_special_tokens=True)
     generate_kwargs = dict(
         {"input_ids": input_ids},
@@ -105,7 +106,7 @@ def generate_fusechat_response():
     return "".join(outputs)
 
 # User-provided prompt
-if prompt := st.chat_input("Hello there! How are you doing?"):
+if prompt := st.chat_input("Do androids dream of electric sheep?"):
     st.session_state.messages.append({"role": "user", "content": prompt})
     with st.chat_message("user"):
         st.write(prompt)