-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathrun_classification.py
178 lines (138 loc) · 8.03 KB
/
run_classification.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
import argparse
import logging
import os
import numpy as np
import sys
from pathlib import Path
from tqdm import tqdm
import pandas as pd
import torch
import torch.nn as nn
import torch.optim as optim
from timm.scheduler import create_scheduler
from timm.optim import create_optimizer
from models import build_model
from utils import train_one_epoch, validation, save_checkpoint
from utils import plot_performance, test_classification, metric_AUROC
from torchmetrics.classification import MultilabelAccuracy
from torchmetrics.functional.classification import multilabel_auroc
def run_experiments(args, train_loader, val_loader, test_loader, model_path, output_path):
accuracy = []
mean_auc_all_runs = []
for idx in range(args.num_trial):
torch.cuda.empty_cache()
model = build_model(args)
model = model.to(args.device)
# optimizer = optim.SGD(model.parameters(), lr=args.base_lr, momentum=0.9)
optimizer = create_optimizer(args, model)
lr_scheduler, _ = create_scheduler(args, optimizer)
loss_fn = nn.BCEWithLogitsLoss()
print (f"Run: {idx+1}\nstart training....")
experiment = args.exp_name + "_run_" + str(idx+1)
save_model_path = model_path.joinpath(experiment)
args.plot_path = output_path / (experiment+ ".pdf")
log_file_train = output_path.joinpath(f"run_{str(idx+1)}.log")
logger1 = logging.getLogger("training_logger")
logger1.setLevel(logging.INFO)
formatter = logging.Formatter("[%(asctime)s.%(msecs)03d] %(message)s",
datefmt="%Y-%m-%d %H:%M:%S")
train_handler = logging.FileHandler(str(log_file_train), mode="a")
train_handler.setFormatter(formatter)
logger1.addHandler(train_handler)
logger1.info(str(args))
loss_train_hist = []
loss_valid_hist = []
acc_train_hist = []
acc_valid_hist = []
best_loss_valid = torch.inf
epoch_counter = 0
patience_counter = 0
for epoch in range(epoch_counter, args.epochs):
model, loss_train, acc_train = train_one_epoch(args,
model,
train_loader,
loss_fn,
optimizer)
logger1.info(f"Run:{idx+1}-Epoch:{epoch+1}, TrainLoss:{loss_train:0.4f}, TrainAcc:{acc_train:0.4f}",
exc_info=True)
loss_train_hist.append(loss_train)
acc_train_hist.append(acc_train)
print("start validation.....")
loss_valid, acc_valid = validation(args, model, val_loader, loss_fn)
logger1.info(f"Run:{idx+1}-Epoch:{epoch+1}, ValidLoss:{loss_valid:0.4f}, ValidAcc:{acc_valid:0.4f}",
exc_info=True)
loss_valid_hist.append(loss_valid)
acc_valid_hist.append(acc_valid)
if loss_valid < best_loss_valid:
save_checkpoint({
'epoch': epoch + 1,
'lossMIN': best_loss_valid,
'state_dict': model.state_dict(),
'optimizer': optimizer.state_dict(),
'scheduler': lr_scheduler.state_dict()
}, filename=str(save_model_path))
best_loss_valid = loss_valid
print('Model Saved!')
patience_counter = 0
else:
print(f"Epoch {epoch+1}: val_loss did not improve from {best_loss_valid}")
patience_counter += 1
if patience_counter > args.patience:
print("Early Stopping")
for handler in logger1.handlers:
handler.close()
logger1.removeHandler(handler)
break
epoch_counter += 1
for handler in logger1.handlers:
handler.close()
logger1.removeHandler(handler)
plot_performance(args, loss_train_hist, loss_valid_hist,
acc_train_hist, acc_valid_hist, epoch_counter)
print ("start testing.....")
log_file_results = output_path.joinpath(f"results_run_{str(idx+1)}.log")
logger2 = logging.getLogger("results_logger")
logger2.setLevel(logging.INFO)
formatter = logging.Formatter("[%(asctime)s.%(msecs)03d] %(message)s",
datefmt="%Y-%m-%d %H:%M:%S")
test_handler = logging.FileHandler(str(log_file_results), mode="a")
test_handler.setFormatter(formatter)
logger2.addHandler(test_handler)
saved_model = model_path.joinpath(f"{experiment}.pth.tar")
y_test, p_test = test_classification(args, str(saved_model), test_loader)
print(y_test.dtype, p_test.dtype, y_test.shape, p_test.shape)
mla_acc_indv = MultilabelAccuracy(num_labels=args.num_classes, average=None).to(args.device)
mla_acc = MultilabelAccuracy(num_labels=args.num_classes).to(args.device)
individual_acc = mla_acc_indv(p_test, y_test).cpu().numpy()
acc = mla_acc(p_test, y_test).cpu().numpy()
print(">>{}: ACC_ClassWise = {}\n".format(experiment, np.array2string(individual_acc, precision=4, separator='\t')))
logger2.info("{}: ACC_ClassWise = {}\n".format(experiment, np.array2string(individual_acc, precision=4, separator='\t')))
print(">>{}: ACC_All_Classes = {}\n".format(experiment, np.array2string(acc, precision=4, separator='\t')))
logger2.info("{}: ACC_All_Classes = {}\n".format(experiment, np.array2string(acc, precision=4, separator='\t')))
auroc_mean = multilabel_auroc(p_test, y_test, num_labels=args.num_classes,
average="macro").cpu().numpy()
auroc_individual = multilabel_auroc(p_test, y_test, num_labels=args.num_classes,
average=None).cpu().numpy()
print(">>{}: AUC_ClassWise = {}".format(experiment, np.array2string(auroc_individual, precision=4, separator=',')))
logger2.info("{}: AUC_ClassWise = {}\n".format(experiment, np.array2string(auroc_individual, precision=4, separator='\t')))
print(">>{}: AUC_All_Classes = {:.4f}".format(experiment, auroc_mean))
logger2.info("{}: AUC_All_Classes = {:.4f}\n".format(experiment, auroc_mean))
accuracy.append(acc.tolist())
mean_auc_all_runs.append(auroc_mean.tolist())
accuracy = np.array(accuracy)
mean_auc_all_runs = np.array(mean_auc_all_runs)
print(">> All trials on all classes: ACC = {}".format(np.array2string(accuracy, precision=4, separator=',')))
logger2.info("All trials on all classes: ACC = {}\n".format(np.array2string(accuracy, precision=4, separator='\t')))
print(">> Mean ACC over All trials: = {:0.4f}".format(np.mean(accuracy)))
logger2.info("Mean ACC over All trials = {:0.4f}\n".format(np.mean(accuracy)))
print(">> ACC_STD over All trials: = {:0.4f}".format(np.std(accuracy)))
logger2.info("ACC_STD over All trials: = {:0.4f}\n".format(np.std(accuracy)))
print(">> All trials on all classes: AUC = {}".format(np.array2string(mean_auc_all_runs, precision=4, separator=',')))
logger2.info("All trials on all classes: AUC = {}\n".format(np.array2string(mean_auc_all_runs, precision=4, separator='\t')))
print(">> Mean AUC over All trials: = {:0.4f}".format(np.mean(mean_auc_all_runs)))
logger2.info("Mean AUC over All trials = {:0.4f}\n".format(np.mean(mean_auc_all_runs)))
print(">> AUC_STD over All trials: = {:0.4f}".format(np.std(mean_auc_all_runs)))
logger2.info("AUC_STD over All trials: = {:0.4f}\n".format(np.std(mean_auc_all_runs)))
for handler in logger2.handlers:
handler.close()
logger2.removeHandler(handler)