Update app.py

app.py

@@ -1,10 +1,6 @@
 #!/usr/bin/env python
 # coding: utf-8
 
-
-# In[48]:
-
-
 import gradio as gr
 from llama_cpp import Llama
 from langchain_community.llms import LlamaCpp
@@ -18,25 +14,39 @@ import pandas as pd
 import re
 import os
 from sklearn.metrics.pairwise import cosine_similarity
+from transformers import AutoTokenizer, AutoModelForCausalLM
+
 
 model = SentenceTransformer('sentence-transformers/paraphrase-multilingual-mpnet-base-v2',device='cpu')
 
 
-llm = LlamaCpp(
-    model_path=r"MediaTek-Research/Breeze-7B-Instruct-v0_1",
-    n_gpu_layers=100,
-    n_batch=512,
-    n_ctx=3000,
-    f16_kv=True,
-    callback_manager=CallbackManager([StreamingStdOutCallbackHandler()]),
-    verbose=False,
-)
+model1 = AutoModelForCausalLM.from_pretrained("MediaTek-Research/Breeze-7B-Instruct-v0_1")
+
+def invoke_with_temperature(prompt, temperature=0.4):
+    # 將 prompt 編碼為模型的輸入格式
+    inputs = tokenizer(prompt, return_tensors="pt")
+    
+    # 使用生成方法生成文本，設置溫度參數
+    output = model1.generate(inputs["input_ids"], max_length=200, temperature=temperature, num_return_sequences=1)
+
+    # 解碼並返回生成的文本
+    return tokenizer.decode(output[0], skip_special_tokens=True)
+    
+# llm = LlamaCpp(
+#     model_path=r"MediaTek-Research/Breeze-7B-Instruct-v0_1",
+#     n_gpu_layers=100,
+#     n_batch=512,
+#     n_ctx=3000,
+#     f16_kv=True,
+#     callback_manager=CallbackManager([StreamingStdOutCallbackHandler()]),
+#     verbose=False,
+# )
 
 embedd_bk=pd.read_pickle(r".\bk_description1_角色形容詞_677.pkl")
 df_bk=pd.read_excel(r".\bk_description1_角色形容詞短文.xlsx")
 
-def invoke_with_temperature(prompt, temperature=0.4):
-    return llm.invoke(prompt, temperature=temperature)
+# def invoke_with_temperature(prompt, temperature=0.4):
+#     return llm.invoke(prompt, temperature=temperature)
 
 def process_user_input(message):
     user_mental_state4= PromptTemplate(