-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathmain_epoch_dnashape.py
204 lines (167 loc) · 10.7 KB
/
main_epoch_dnashape.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
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
#!/usr/bin/env python
# -*- encoding: utf-8 -*-
'''
@File : main.py
@Time : 2023/11/09 11:18:29
@Author : Cmf
@Version : 1.0
@Desc : None
'''
# here put the import lib
import click
import numpy as np
from functools import partial
from pathlib import Path
from ruamel.yaml import YAML
from sklearn.model_selection import train_test_split
from torch.utils.data.dataloader import DataLoader
from logzero import logger
from tfnet.data_utils import *
from tfnet.datasets_dnashape import TFBindDataset
from tfnet.models_dnashape_epoch import Model
from tfnet.evaluation import output_eval, output_predict, CUTOFF
import pdb
# code
def train(model, data_cnf, model_cnf, train_data, valid_data=None, class_weights_dict = None, random_state=1240):
logger.info(f'Start training model {model.model_path}')
if valid_data is None:
train_data, valid_data = train_test_split(train_data, test_size=data_cnf.get('valid', 0.2),
random_state=random_state)
model.train(data_cnf, model_cnf, train_data, valid_data, class_weights_dict, **model_cnf['train']) # for samples_per_epoch
logger.info(f'Finish training model {model.model_path}')
def test(model, data_cnf, model_cnf, test_data):
data_loader = DataLoader(TFBindDataset(test_data, data_cnf['genome_fasta_file'], data_cnf['bigwig_file'], **model_cnf['padding']),
batch_size=model_cnf['test']['batch_size'])
return model.predict(data_loader)
def generate_cv_id(length, num_groups=5):
base_size = length // num_groups
extra_size = length % num_groups
group_sizes = [base_size + 1 if i < extra_size else base_size for i in range(num_groups)]
#labels = np.repeat(np.arange(1, num_groups + 1), group_sizes)
labels = np.concatenate([np.full(size, i) for i, size in enumerate(group_sizes)])
return labels
@click.command()
@click.option('-d', '--data-cnf', type=click.Path(exists=True))
@click.option('-m', '--model-cnf', type=click.Path(exists=True))
@click.option('--mode', type=click.Choice(('train', 'eval', 'predict','5cv', 'loo', 'lomo')), default=None)
@click.option('-s', '--start-id', default=0)
@click.option('-n', '--num_models', default=1)
@click.option('-c', '--continue_train', is_flag=True)
def main(data_cnf, model_cnf, mode, start_id, num_models, continue_train):
yaml = YAML(typ='safe')
data_cnf, model_cnf = yaml.load(Path(data_cnf)), yaml.load(Path(model_cnf))
logger.info(f'check important parameter')
logger.info(f"Model Name: {model_cnf['name']}, tf number: {len(model_cnf['model']['all_tfs'])}, cutoff: {CUTOFF}")
model_name = model_cnf['name']
model_path = Path(model_cnf['path'])/f'{model_name}.pt'
res_path = Path(data_cnf['results'])/f'{model_name}'
Path(data_cnf['results']).mkdir(parents=True, exist_ok=True)
model_cnf.setdefault('ensemble', 20)
get_data_fn = partial(get_data_lazy, genome_fasta_file= data_cnf['genome_fasta_file'])
all_tfs = model_cnf['model']['all_tfs']
model_structure = model_cnf['model_structure']
if model_structure == 'Danq':
from tfnet.networks_danq import Danq as Selected_Model
elif model_structure == 'scFAN':
from tfnet.networks_scfan import scFAN as Selected_Model
elif model_structure == 'DeepATT':
from tfnet.networks_deepatt import DeepATT as Selected_Model
elif model_structure == 'DeepFormer':
from tfnet.networks_deepformer import DeepFormer as Selected_Model
elif model_structure == 'TFNet3':
from tfnet.networks_tfnet3 import TFNet3 as Selected_Model
else:
raise ValueError(f"Unknown network type: {model_structure}")
classweights = model_cnf['classweights']
if classweights and ( mode != "eval" and model != "eval_list" and model != "predict"):
class_weights_dict = calculate_class_weights_dict(data_cnf['train'])
else :
class_weights_dict = None
if mode == "train":
for model_id in range(start_id, start_id + num_models):
if continue_train:
logger.info(f'Continue train Mode')
model = Model(Selected_Model, model_path=model_path.with_stem(f'{model_path.stem}-{model_id}'), class_weights_dict = class_weights_dict,
**model_cnf['model'])
logger.info(f'Loading Model: {model_path.stem}-{model_id}')
model.load_model()
train_data = get_data_fn(data_cnf['train']) if mode is None or mode == 'train' else None
valid_data = get_data_fn(data_cnf['valid']) if train_data is not None and 'valid' in data_cnf else None
train(model, data_cnf, model_cnf, train_data=train_data, valid_data=valid_data, class_weights_dict = class_weights_dict)
else:
if not model_path.with_stem(f'{model_path.stem}-{model_id}').exists():
model = Model(Selected_Model, model_path=model_path.with_stem(f'{model_path.stem}-{model_id}'), class_weights_dict = class_weights_dict,
**model_cnf['model'])
train_data = get_data_fn(data_cnf['train']) if mode is None or mode == 'train' else None
valid_data = get_data_fn(data_cnf['valid']) if train_data is not None and 'valid' in data_cnf else None
train(model, data_cnf, model_cnf, train_data=train_data, valid_data=valid_data, class_weights_dict = class_weights_dict)
else:
logger.info(f'Model already exsit: {model_path.stem}-{model_id}')
elif mode == 'eval':
test_data = get_data_fn(data_cnf['test'])
shift = int((model_cnf['padding']['DNA_len'] - model_cnf['padding']['target_len'])/2)
#chr, start, stop, targets_lists = [x[0] for x in test_data], [x[1] + shift for x in test_data], [x[2] - shift for x in test_data], [x[-2] for x in test_data] # depend on the input data len
chr, start, stop, targets_lists = [x[0] for x in test_data], [x[1] for x in test_data], [x[2] for x in test_data], [x[-1] for x in test_data]
scores_lists = []
for model_id in range(start_id, start_id + num_models):
model = Model(Selected_Model, model_path=model_path.with_stem(f'{model_path.stem}-{model_id}'), class_weights_dict = class_weights_dict,
**model_cnf['model'])
scores_lists.append(test(model, data_cnf, model_cnf, test_data=test_data))
output_eval(chr, start, stop, np.array(targets_lists), np.mean(scores_lists, axis=0), all_tfs, res_path)
elif mode == 'predict':
predict_data = get_data_fn(data_cnf['predict'])
shift = int((model_cnf['padding']['DNA_len'] - model_cnf['padding']['target_len'])/2)
chr, start, stop, targets_lists = [x[0] for x in predict_data], [x[1] + shift for x in predict_data], [x[2] - shift for x in predict_data], [x[-1] for x in predict_data]
scores_lists = []
for model_id in range(start_id, start_id + num_models):
model = Model(Selected_Model, model_path=model_path.with_stem(f'{model_path.stem}-{model_id}'), class_weights_dict = class_weights_dict,
**model_cnf['model'])
scores_lists.append(test(model, data_cnf, model_cnf, test_data=predict_data))
output_predict(chr, start, stop, np.mean(scores_lists, axis=0), res_path)
elif mode == '5cv':
data = np.asarray(get_data_fn(data_cnf['train']), dtype=object)
data_group_name, atac_signal, data_truth = [x[0] for x in data], [x[1] for x in data], [x[2] for x in data]
cv_id_len = data.shape[0]
# ---------------------- generate cv id for use rather than read a input ---------------------- #
cv_id = generate_cv_id(cv_id_len)
assert len(data) == len(cv_id)
scores_list = []
for model_id in range(start_id, start_id + num_models):
scores_ = np.empty(len(data)*len(all_tfs), dtype=np.float32).reshape(len(data), len(all_tfs))
for cv_ in range(5):
train_data, test_data = data[cv_id != cv_], data[cv_id == cv_]
model = Model(Selected_Model, model_path=model_path.with_stem(f'{model_path.stem}-{model_id}-CV{cv_}'), class_weights_dict = class_weights_dict,
**model_cnf['model'])
if not continue_train or not model.model_path.exists():
train(model, data_cnf, model_cnf, train_data=train_data, class_weights_dict = class_weights_dict)
scores_[cv_id == cv_] = test(model, model_cnf, test_data=test_data)
scores_list.append(scores_)
#pdb.set_trace()
#output_res(np.array(data_group_name)[cv_id == cv_], np.array(data_truth)[cv_id == cv_], np.mean(scores_[cv_id == cv_], axis=0),
output_eval(np.array(data_group_name)[cv_id == cv_], np.array(data_truth)[cv_id == cv_], scores_[cv_id == cv_], all_tfs,
res_path.with_name(f'{res_path.stem}-5CV'))
elif mode == 'loo' or mode == 'lomo':
data = np.asarray(get_data_fn(data_cnf['train']), dtype=object)
with open(data_cnf['cv_id']) as fp:
cv_id = np.asarray([int(line) for line in fp])
scores_list = []
for model_id in range(start_id, start_id + num_models):
group_names, group_names_, truth_, scores_ = np.asarray([x[0] for x in data]), [], [], []
for name_ in sorted(set(group_names)):
train_data, train_cv_id = data[group_names != name_], cv_id[group_names != name_]
test_data, test_cv_id = data[group_names == name_], cv_id[group_names == name_]
if len(test_data) > 30 and len([x[-1] for x in test_data if x[-1] >= CUTOFF]) >= 3:
for cv_ in range(5):
model = Model(Selected_Model,
model_path=model_path.with_stem(F'{model_path.stem}-{name_}-{model_id}-CV{cv_}'), class_weights_dict = class_weights_dict,
**model_cnf['model'])
if not model.model_path.exists() or not continue_train:
train(model, data_cnf, model_cnf, train_data[train_cv_id != cv_], class_weights_dict=class_weights_dict)
test_data_ = test_data[test_cv_id == cv_]
group_names_ += [x[0] for x in test_data_]
truth_ += [x[-1] for x in test_data_]
scores_ += test(model, model_cnf, test_data_).tolist()
scores_list.append(scores_)
output_eval(group_names_, truth_, np.mean(scores_list, axis=0), all_tfs, res_path.with_name(f'{res_path.stem}-LOMO'))
if __name__ == '__main__':
main()