-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathsequence_processing_longformer.py
131 lines (103 loc) · 5.28 KB
/
sequence_processing_longformer.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
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
import numpy as np
from transformers import LongformerModel, LongformerTokenizer, LongformerConfig
# from transformers import pipeline # !!! can not import pipline.......
print("test!")
def get_max_length(seq_list, tokenizer):
max_value = 0
for txt in seq_list:
tokens = tokenizer.tokenize(txt)
token_value = len(tokens)
# print(tokens)
# print(token_value)
if token_value > max_value:
max_value = token_value
print("max_length is:", max_value)
return max_value
def get_data_max_length(enhancers, promoters):
ep_list = [enhancer + tokenizer.sep_token + promoter for enhancer, promoter in zip(enhancers, promoters)]
max_length = get_max_length(ep_list, tokenizer)
print("MAX:", max_length, "##" * 50) # 2626 2659
return max_length
def get_data(enhancers, promoters, tokenizer, model, max_length):
X_enprs = []
for enhancer, promoter in zip(enhancers, promoters):
encoded_inputs = tokenizer(enhancer + tokenizer.sep_token + promoter, return_tensors='pt', padding=True,
max_length=max_length)
X_enpr_tensor = model(**encoded_inputs)[0]
# print(X_enpr_features.shape)
X_enpr_np = X_enpr_tensor.detach().numpy()
X_enprs.append(X_enpr_np)
# ep_list = [enhancer + tokenizer.sep_token + promoter for enhancer, promoter in zip(enhancers, promoters)]
# encoded_inputs = tokenizer(ep_list, return_tensors='pt', padding=True)
# X_enpr_features = model(**encoded_inputs, return_netsors='pt')
print("add:", len(X_enprs))
return X_enprs
def save_data(enhancers, promoters, tokenizer, model, max_length, file):
with open(file, 'ab') as f:
for enhancer, promoter in zip(enhancers, promoters):
print(tokenizer.sep_token)
encoded_inputs = tokenizer(enhancer + tokenizer.sep_token + promoter, return_tensors='pt', padding=True,
max_length=max_length)
X_enpr_tensor = model(**encoded_inputs)[0]
# print(X_enpr_features.shape)
X_enpr_np = X_enpr_tensor.detach().numpy()
for item in X_enpr_np:
# print(item)
np.savetxt(f, item, delimiter=',')
f.close()
# ep_list = [enhancer + tokenizer.sep_token + promoter for enhancer, promoter in zip(enhancers, promoters)]
# encoded_inputs = tokenizer(ep_list, return_tensors='pt', padding=True)
# X_enpr_features = model(**encoded_inputs, return_netsors='pt')
print("save over!")
# In[]:
names = ['pbc_IMR90', 'GM12878', 'HUVEC', 'HeLa-S3', 'IMR90', 'K562', 'NHEK']
name = names[0]
feature_name = "longformer-hug"
train_dir = 'data/%s/train/' % name
imbltrain = 'data/%s/imbltrain/' % name
test_dir = 'data/%s/test/' % name
Data_dir = 'data/%s/%s/' % (name, feature_name)
print('Experiment on %s dataset' % name)
print('Loading seq data...')
enhancers_tra = open(train_dir + '%s_enhancer.fasta' % name, 'r').read().splitlines()[1::2]
promoters_tra = open(train_dir + '%s_promoter.fasta' % name, 'r').read().splitlines()[1::2]
y_tra = np.loadtxt(train_dir + '%s_label.txt' % name)
im_enhancers_tra = open(imbltrain + '%s_enhancer.fasta' % name, 'r').read().splitlines()[1::2]
im_promoters_tra = open(imbltrain + '%s_promoter.fasta' % name, 'r').read().splitlines()[1::2]
y_imtra = np.loadtxt(imbltrain + '%s_label.txt' % name)
enhancers_tes = open(test_dir + '%s_enhancer_test.fasta' % name, 'r').read().splitlines()[1::2]
promoters_tes = open(test_dir + '%s_promoter_test.fasta' % name, 'r').read().splitlines()[1::2]
y_tes = np.loadtxt(test_dir + '%s_label_test.txt' % name)
print('平衡训练集')
print('pos_samples:' + str(int(sum(y_tra))))
print('neg_samples:' + str(len(y_tra) - int(sum(y_tra))))
print('不平衡训练集')
print('pos_samples:' + str(int(sum(y_imtra))))
print('neg_samples:' + str(len(y_imtra) - int(sum(y_imtra))))
print('测试集')
print('pos_samples:' + str(int(sum(y_tes))))
print('neg_samples:' + str(len(y_tes) - int(sum(y_tes))))
# In[ ]:
model_name = 'pre-model/' + 'longformer-base-4096'
config = LongformerConfig.from_pretrained(model_name)
tokenizer = LongformerTokenizer.from_pretrained(model_name)
model = LongformerModel.from_pretrained(model_name, config=config)
a1 = get_data_max_length(enhancers_tra, promoters_tra)
a2 = get_data_max_length(im_enhancers_tra, im_promoters_tra)
a3 = get_data_max_length(enhancers_tes, promoters_tes)
max_value = max(a1, a2, a3)
print(max_value)
# X_enpr_tra = get_data(enhancers_tra, promoters_tra, tokenizer, model)
# np.savez(Data_dir + '%s_train.npz' % name, X_enpr_tra=X_enpr_tra, y_tra=y_tra)
# print("saved train!")
# save_data(enhancers_tra, promoters_tra, tokenizer, model, max_value, Data_dir + '%s_train.csv' % name)
#
# X_enpr_imtra = get_data(im_enhancers_tra, im_promoters_tra, tokenizer, model)
# np.savez(Data_dir + 'im_%s_train.npz' % name, X_enpr_tra=X_enpr_imtra, y_tra=y_imtra)
# print("saved im_train!")
# save_data(im_enhancers_tra, im_promoters_tra, tokenizer, model, max_value, Data_dir + 'im_%s_train.csv' % name)
#
# X_enpr_tes = get_data(enhancers_tes, promoters_tes, tokenizer, model)
# np.savez(Data_dir + '%s_test.npz' % name, X_enpr_tes=X_enpr_tes, y_tes=y_tes)
# print("saved test!")
# save_data(enhancers_tes, promoters_tes, tokenizer, model, max_value, Data_dir + '%s_test.csv' % name)