ap2p.py

# import streamlit as st
# import torch
# import transformers
# from transformers import pipeline
# from transformers import LlamaTokenizer, LlamaForCausalLM
# import time
# import csv
# import locale
# locale.getpreferredencoding = lambda: "UTF-8"




# -




# #https://huggingface.co/shibing624/chinese-alpaca-plus-7b-hf
# #https://huggingface.co/ziqingyang/chinese-alpaca-2-7b
# #https://huggingface.co/minlik/chinese-alpaca-plus-7b-merged

# def generate_prompt(text):
#     return f"""Below is an instruction that describes a task. Write a response that appropriately completes the request.
#     ### Instruction:
#     {text}
    
#     ### Response:"""

# tokenizer = LlamaTokenizer.from_pretrained('shibing624/chinese-alpaca-plus-7b-hf')
# pipeline = pipeline(
#     "text-generation",
#     model="shibing624/chinese-alpaca-plus-7b-hf",
#     torch_dtype=torch.float32,
#     device_map="auto",
# )

# st.title("Chinese text generation alpaca2")
# st.write("Enter a sentence and alpaca2 will answer:")

# user_input = st.text_input("")




# with open('alpaca_output.csv', 'a', newline='',encoding = "utf-8") as csvfile:
#     writer = csv.writer(csvfile)
#     # writer.writerow(["stockname",'prompt','answer','time'])
#     if user_input:
#         if user_input[0] == ".":
#             stockname = user_input[1:4]
#             analysis = user_input[4:]
            
#             text = f"""請以肯定和專業的語氣，一步一步的思考並回答以下關於{stockname}的問題，避免空洞的答覆:
#             - 請回答關於{stockname}的問題，請總結給予的資料以及資料解釋，並整合出金融上的洞見。\n
#             - 請不要生成任何資料沒有提供的數據，即便你已知道。\n
#             - 請假裝這些資料都是你預先知道的知識。因此，請不要提到「根據資料」、「基於上述資料」等回答
#             - 請不要說「好的、我明白了、根據我的要求、以下是我的答案」等贅詞，請輸出分析結果即可\n
#             - 請寫300字到500字之間，若合適，可以進行分類、列點
#             資料:{stockname}{analysis}
            
#             請特別注意，分析結果包含籌碼面、基本面以及技術面，請針對這三個面向進行回答，並且特別注意個別符合幾項和不符合幾項。籌碼面、技術面和基本面滿分十分，總計滿分為30分。
#             三個面向中，符合5項以上代表該面項表現好，反之是該面項表現差。
#             """
        
#             prompt = generate_prompt(text)
#             start = time.time()
#             sequences = pipeline(
#                 prompt,
#                 do_sample=True,
#                 top_k=40,
#                 num_return_sequences=1,
#                 eos_token_id=tokenizer.eos_token_id,
#                 max_length=200,
#                 )
#             end = time.time()
#             for seq in sequences:
#                 st.write(f"Result: {seq}") #seq['generated_text']
#             st.write(f"time: {(end-start):.2f}")
#             writer.writerow([stockname,text,sequences,f"time: {(end-start):.2f}"])
        
#             # input_ids = tokenizer.encode(prompt, return_tensors='pt').to('cuda')
#             # with torch.no_grad():
#             #     output_ids = model.generate(
#             #         input_ids=input_ids,
#             #         max_new_tokens=2048,
#             #         top_k=40,
        
#             #     ).cuda()
#             # output = tokenizer.decode(output_ids[0], skip_special_tokens=True)
#         else:
#             prompt = generate_prompt(user_input)
#             start = time.time()
#             sequences = pipeline(
#                 prompt,
#                 do_sample=True,
#                 top_k=40,
#                 num_return_sequences=1,
#                 eos_token_id=tokenizer.eos_token_id,
#                 max_length=200,
#                 )
#             end = time.time()
#             for seq in sequences:
#                 st.write(f"Result: {seq}") #seq['generated_text']
#             st.write(f"time: {(end-start):.2f}")
#             writer.writerow(["無",user_input,sequences,f"time: {(end-start):.2f}"])