-
Notifications
You must be signed in to change notification settings - Fork 0
/
eval.py
160 lines (134 loc) · 5.76 KB
/
eval.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
# Copyright (c) Microsoft Corporation.
# Licensed under the MIT License.
import matplotlib.pyplot as plt
import os
import json
import numpy as np
import torch
from torch.utils.data import DataLoader
from semilearn.core.utils import get_net_builder, get_dataset
from metrics import expected_calibration_error, ECELoss, ClasswiseECELoss, BrierScore, AdaptiveECELoss
import torch
import numpy as np
from sklearn.manifold import TSNE
import matplotlib.pyplot as plt
from sklearn.metrics import confusion_matrix
import seaborn as sns
if __name__ == "__main__":
import argparse
parser = argparse.ArgumentParser()
parser.add_argument('--load_path', type=str, required=True)
'''
Backbone Net Configurations
'''
parser.add_argument('--net', type=str, default='vit_small_patch2_32')
parser.add_argument('--net_from_name', type=bool, default=False)
'''
Data Configurations
'''
parser.add_argument('--batch_size', type=int, default=16)
parser.add_argument('--data_dir', type=str, default='./data')
parser.add_argument('--dataset', type=str, default='cifar100')
parser.add_argument('--num_classes', type=int, default=100)
parser.add_argument('--img_size', type=int, default=32)
parser.add_argument('--crop_ratio', type=int, default=0.875)
parser.add_argument('--max_length', type=int, default=512)
parser.add_argument('--max_length_seconds', type=float, default=4.0)
parser.add_argument('--sample_rate', type=int, default=16000)
args = parser.parse_args()
checkpoint_path = os.path.join(args.load_path)
checkpoint = torch.load(checkpoint_path)
load_model = checkpoint['ema_model']
load_state_dict = {}
for key, item in load_model.items():
if key.startswith('module'):
new_key = '.'.join(key.split('.')[1:])
load_state_dict[new_key] = item
else:
load_state_dict[key] = item
save_dir = '/'.join(checkpoint_path.split('/')[:-1])
args.save_dir = save_dir
args.save_name = ''
net = get_net_builder(args.net, args.net_from_name)(num_classes=args.num_classes)
keys = net.load_state_dict(load_state_dict)
if torch.cuda.is_available():
net.cuda()
net.eval()
# specify these arguments manually
args.num_labels = 200
args.ulb_num_labels = 49600
args.lb_imb_ratio = 1
args.ulb_imb_ratio = 1
args.seed = 0
args.epoch = 1
args.num_train_iter = 1024
dataset_dict = get_dataset(args, 'fixmatch', args.dataset, args.num_labels, args.num_classes, args.data_dir, False)
eval_dset = dataset_dict['eval']
eval_loader = DataLoader(eval_dset, batch_size=args.batch_size, drop_last=False, shuffle=False, num_workers=4)
print('LENGTH OF TEST SET : ', len(eval_dset))
acc = 0.0
ece = 0.0
aece = 0.0
cece = 0.0
brier = 0.0
test_feats = []
test_preds = []
test_probs = []
test_labels = []
metrics_dict = {}
checkpoint_file_name = os.path.basename(args.load_path)
checkpoint_directory = os.path.dirname(args.load_path)
with torch.no_grad():
ece_criterion = ECELoss(15).cuda()
classwise_ece_loss = ClasswiseECELoss(n_bins=15).cuda()
adaptive_ece_loss = AdaptiveECELoss(n_bins=15).cuda()
brier_loss = BrierScore()
accuracy_values = [] # List to store accuracy values (0 or 1)
confidence_values = [] # List to store confidence (probability) values
for data in eval_loader:
image = data['x_lb']
target = data['y_lb'].cpu() # Move target to CPU
image = image.type(torch.FloatTensor).cuda()
feat = net(image, only_feat=True)
logit = net(feat, only_fc=True)
prob = logit.softmax(dim=-1)
pred = prob.argmax(1)
accuracy = pred.cpu().eq(target).numpy()
acc += accuracy.sum()
confidence = prob.cpu().numpy()
accuracy_values.extend(accuracy)
confidence_values.extend(confidence)
test_feats.append(feat.cpu().numpy())
test_preds.append(pred.cpu().numpy())
test_probs.append(logit.cpu())
test_labels.append(target)
test_feats = np.concatenate(test_feats)
test_preds = np.concatenate(test_preds)
test_probs = np.concatenate(test_probs)
test_labels = np.concatenate(test_labels)
ece = ece_criterion(torch.from_numpy(test_probs), torch.from_numpy(test_labels))
aece = adaptive_ece_loss(torch.from_numpy(test_probs), torch.from_numpy(test_labels))
cece = classwise_ece_loss(torch.from_numpy(test_probs), torch.from_numpy(test_labels))
brier = brier_loss(torch.from_numpy(test_probs), torch.from_numpy(test_labels))
metrics_dict['Test ECE'] = ece
metrics_dict['Test CECE'] = cece
metrics_dict['Test AECE'] = aece
metrics_dict['Test Brier'] = brier
metrics_dict['Test Accuracy'] = acc / len(eval_dset)
metrics_dict['Test Error'] = 1 - acc / len(eval_dset)
print("Metrics Dictionary:", metrics_dict)
# Convert tensor types to lists or numbers
converted_dict = {}
for key, value in metrics_dict.items():
if isinstance(value, torch.Tensor):
if value.dim() == 0:
converted_dict[key] = value.item() # Convert single value tensor to number
else:
converted_dict[key] = value.tolist() # Convert tensor to a list
checkpoint_file_name = os.path.basename(args.load_path)
checkpoint_directory = os.path.dirname(args.load_path)
metrics_file_name = os.path.splitext(checkpoint_file_name)[0] + "_metrics.json"
metrics_file_path = os.path.join(checkpoint_directory, metrics_file_name)
with open(metrics_file_path, 'w') as metrics_file:
json.dump(converted_dict, metrics_file)
print(f"Metrics saved to: {metrics_file_path}")