-
Notifications
You must be signed in to change notification settings - Fork 0
/
utils.py
104 lines (87 loc) · 4.62 KB
/
utils.py
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
import os
import json
from typing import List, Optional
from sklearn.metrics import confusion_matrix, f1_score
# this file stores all utility functions, such as calculating the metrics, preprocessing the datasets
class Dataset:
def __init__(self, texts: List[str], label_names: List[str], prompt: Optional[str] = None):
self.texts = texts
self.label_names = label_names
self.prompt = prompt
class Labels:
def __init__(self, labels: List[int]):
self.labels = labels
def data_process(dataset_name, label_name_dir, prompt_dir):
train_text_path = os.path.join("data", dataset_name, "train_text.txt")
with open(train_text_path, mode='r', encoding='utf-8') as file:
train_text = list(map(lambda x: x.strip(), file.readlines()))
label_names_path = label_name_dir #os.path.join("data", dataset_name, "label_names.txt")
with open(label_names_path, mode='r', encoding='utf-8') as file:
label_names = list(map(lambda x: x.strip(), file.readlines()))
test_text_path = os.path.join("data", dataset_name, "test_text.txt")
with open(test_text_path, mode='r', encoding='utf-8') as file:
test_text = list(map(lambda x: x.strip(), file.readlines()))
test_label_path = os.path.join("data", dataset_name, "test_label.txt")
with open(test_label_path, mode='r', encoding='utf-8') as file:
test_label = list(map(lambda x: int(x.strip()), file.readlines()))
train_label_path = os.path.join("data", dataset_name, "train_label.txt")
train_label = None
if os.path.exists(train_label_path):
with open(train_label_path, mode='r', encoding='utf-8') as file:
train_label = list(map(lambda x: int(x.strip()), file.readlines()))
prompt_path = prompt_dir # os.path.join("data", dataset_name, "prompt.txt")
prompt = None
if os.path.exists(prompt_path):
with open(prompt_path, mode='r', encoding='utf-8') as file:
prompt = file.read()
train_Dataset = Dataset(train_text, label_names, prompt)
train_Labels = Labels(train_label)
test_Dataset = Dataset(test_text, label_names, prompt)
test_Labels = Labels(test_label)
return train_Dataset, train_Labels, test_Dataset, test_Labels
def get_method(method_name, hyperparameter_file_path, base_model):
if method_name == "xclass":
from methods.xclass_method import xclass, xclassHyperparams
hyperparameters = xclassHyperparams.from_dict(json.load(open(hyperparameter_file_path, mode='r')))
method = xclass(hyperparameters, base_model)
elif method_name == "prompt":
from methods.prompt_method import prompt, promptHyperparams
hyperparameters = promptHyperparams.from_dict(json.load(open(hyperparameter_file_path, mode='r')))
method = prompt(hyperparameters, base_model)
elif method_name == "prompt_gpt":
from methods.prompt_gpt_method import prompt_gpt, prompt_gptHyperparams
hyperparameters = prompt_gptHyperparams.from_dict(json.load(open(hyperparameter_file_path, mode='r')))
method = prompt_gpt(hyperparameters, base_model)
elif method_name == "lotclass":
from methods.lotclass_method import lotclass, lotclassHyperparams
hyperparameters = lotclassHyperparams.from_dict(json.load(open(hyperparameter_file_path, mode='r')))
method = lotclass(hyperparameters, base_model)
elif method_name == "npprompt":
from methods.npprompt_method import npprompt, nppromptHyperparams
hyperparameters = nppromptHyperparams.from_dict(json.load(open(hyperparameter_file_path, mode='r')))
method = npprompt(hyperparameters, base_model)
elif method_name == "classkg":
from methods.classkg_method import classkg, classkgHyperparams
hyperparameters = classkgHyperparams.from_dict(json.load(open(hyperparameter_file_path, mode='r')))
method = classkg(hyperparameters, base_model)
else:
... # add new method here
return method
def evaluate_predictions(true_class, predicted_class, output_to_console=True, return_tuple=False):
confusion = confusion_matrix(true_class, predicted_class)
if output_to_console:
print("-" * 80 + "Evaluating" + "-" * 80)
print(confusion)
f1_macro = f1_score(true_class, predicted_class, average='macro')
f1_micro = f1_score(true_class, predicted_class, average='micro')
if output_to_console:
print("F1 macro: " + str(f1_macro))
print("F1 micro: " + str(f1_micro))
if return_tuple:
return confusion, f1_macro, f1_micro
else:
return {
"confusion": confusion.tolist(),
"f1_macro": f1_macro,
"f1_micro": f1_micro
}