1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
| import os
import time
import pandas as pd
from langchain.callbacks.manager import CallbackManager
from langchain.callbacks.streaming_stdout import StreamingStdOutCallbackHandler
from langchain.chains import LLMChain
from langchain.llms import LlamaCpp
from langchain.prompts import PromptTemplate
from tqdm import tqdm
MIN_INTERVAL = 1.5
keys = dict(
title='title',
abstract='abs',
category='cat',
subcategory='subcat',
)
current_path = os.getcwd()
news_df = pd.read_csv(
filepath_or_buffer=os.path.join(current_path + '/data/news.tsv'),
sep='\t',
header=0,
)
news_list = []
for news in tqdm(news_df.iterrows()):
dic = {}
for key in keys:
dic[key] = news[1][keys[key]]
news_list.append((news[1]['nid'], dic))
prompt_template = """You are asked to act as a news title enhancer. I will provide you a piece of news, with its original title, category, subcategory, and abstract (if exists). The news format is as below:
[title] {title}
[abstract] {abstract}
[category] {category}
[subcategory] {subcategory}
where title, abstract, category, and subcategory in the brace will be filled with content. You can only response a rephrased news title which should be clear, complete, objective and neutral. You can expand the title according to the above requirements. You are not allowed to response any other words for any explanation. Your response format should be:
[newtitle]
where [newtitle] should be filled with the enhanced title. Now, your role of news title enhancer formally begins. Any other information should not disturb your role."""
save_path = current_path + '/output/news_summarizer.log'
callback_manager = CallbackManager([StreamingStdOutCallbackHandler()])
llm = LlamaCpp(
model_path="/Users/liuqiang/Desktop/code/llm/models/gguf/qwen1.5-72b-chat-q5_k_m.gguf",
temperature=0.8,
top_p=0.8,
n_ctx=6000,
callback_manager=callback_manager,
verbose=True,
)
"""
PromptTemplate 与 chain的使用案例：
（1）文本的使用方式
prompt = PromptTemplate(
input_variables=["product"],
template="What is a good name for a company that makes {product}?",
)
chain = LLMChain(llm=llm, prompt=prompt)
# Run the chain only specifying the input variable.
print(chain.run("colorful socks"))
（2）字典的使用方式
prompt = PromptTemplate(
input_variables=["company", "product"],
template="What is a good name for {company} that makes {product}?",
)
chain = LLMChain(llm=llm, prompt=prompt)
print(chain.run({
'company': "ABC Startup",
'product': "colorful socks"
}))
"""
prompt = PromptTemplate(
input_variables=["title", "abstract", "category", "subcategory"],
template=prompt_template,
)
chain = LLMChain(llm=llm, prompt=prompt)
exist_set = set()
with open(save_path, 'r') as f:
for line in f:
if line and line.startswith('N'):
exist_set.add(line.split('\t')[0])
for nid, content in tqdm(news_list):
start_time = time.time()
if nid in exist_set:
continue
try:
title = content['title']
abstract = content['abstract']
category = content['category']
subcategory = content['subcategory']
enhanced = chain.run(title=title, abstract=abstract, category=category, subcategory=subcategory)
enhanced = enhanced.rstrip('\n')
with open(save_path, 'a') as f:
f.write(f'{nid}\t{enhanced}\n')
except Exception as e:
print(e)
interval = time.time() - start_time
if interval <= MIN_INTERVAL:
time.sleep(MIN_INTERVAL - interval)
|
微信支付
支付宝
