-
Notifications
You must be signed in to change notification settings - Fork 25
/
utils.py
199 lines (172 loc) · 6.83 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
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
# Copyright (c) [2012]-[2021] YUAN Li@NUS.
#
# This source code is licensed under the Clear BSD License
# LICENSE file in the root directory of this file
# All rights reserved.
'''
- load_for_transfer_learning: load pretrained paramters to model in transfer learning
- get_mean_and_std: calculate the mean and std value of dataset.
- msr_init: net parameter initialization.
- progress_bar: progress bar mimic xlua.progress.
'''
import os
import sys
import time
import torch
import torch.nn as nn
import torch.nn.init as init
import logging
import os
from collections import OrderedDict
_logger = logging.getLogger(__name__)
def resize_pos_embed(posemb, posemb_new): # example: 224:(14x14+1)-> 384: (24x24+1)
# Rescale the grid of position embeddings when loading from state_dict. Adapted from
# https://github.com/google-research/vision_transformer/blob/00883dd691c63a6830751563748663526e811cee/vit_jax/checkpoint.py#L224
ntok_new = posemb_new.shape[1]
if True:
posemb_tok, posemb_grid = posemb[:, :1], posemb[0, 1:] # posemb_tok is for cls token, posemb_grid for the following tokens
ntok_new -= 1
else:
posemb_tok, posemb_grid = posemb[:, :0], posemb[0]
gs_old = int(math.sqrt(len(posemb_grid))) # 14
gs_new = int(math.sqrt(ntok_new)) # 24
_logger.info('Position embedding grid-size from %s to %s', gs_old, gs_new)
posemb_grid = posemb_grid.reshape(1, gs_old, gs_old, -1).permute(0, 3, 1, 2) # [1, 196, dim]->[1, 14, 14, dim]->[1, dim, 14, 14]
posemb_grid = F.interpolate(posemb_grid, size=(gs_new, gs_new), mode='bicubic') # [1, dim, 14, 14] -> [1, dim, 24, 24]
posemb_grid = posemb_grid.permute(0, 2, 3, 1).reshape(1, gs_new * gs_new, -1) # [1, dim, 24, 24] -> [1, 24*24, dim]
posemb = torch.cat([posemb_tok, posemb_grid], dim=1) # [1, 24*24+1, dim]
return posemb
def load_state_dict(checkpoint_path, model, num_classes, use_ema=False, del_posemb=False):
if checkpoint_path and os.path.isfile(checkpoint_path):
checkpoint = torch.load(checkpoint_path, map_location='cpu')
state_dict_key = 'state_dict'
if isinstance(checkpoint, dict):
if use_ema and 'state_dict_ema' in checkpoint:
state_dict_key = 'state_dict_ema'
if state_dict_key and state_dict_key in checkpoint:
new_state_dict = OrderedDict()
for k, v in checkpoint[state_dict_key].items():
# strip `module.` prefix
name = k[7:] if k.startswith('module') else k
new_state_dict[name] = v
state_dict = new_state_dict
else:
state_dict = checkpoint
_logger.info("Loaded {} from checkpoint '{}'".format(state_dict_key, checkpoint_path))
print(f'num classes: {num_classes}')
if num_classes != 1000:
# completely discard fully connected for all other differences between pretrained and created model
print('delete the original class')
del state_dict['head' + '.weight']
del state_dict['head' + '.bias']
if del_posemb==True:
del state_dict['pos_embed']
#old_posemb = state_dict['pos_embed']
#if model.pos_embed.shape != old_posemb.shape: # need resize the position embedding by interpolate
# new_posemb = resize_pos_embed(old_posemb, model.pos_embed)
# state_dict['pos_embed'] = new_posemb
return state_dict
else:
_logger.error("No checkpoint found at '{}'".format(checkpoint_path))
raise FileNotFoundError()
def load_for_transfer_learning(model, checkpoint_path, num_classes, use_ema=False, strict=True):
print(f'num classes: {num_classes}')
state_dict = load_state_dict(checkpoint_path, use_ema, num_classes)
model.load_state_dict(state_dict, strict=strict)
def get_mean_and_std(dataset):
'''Compute the mean and std value of dataset.'''
dataloader = torch.utils.data.DataLoader(dataset, batch_size=1, shuffle=True, num_workers=2)
mean = torch.zeros(3)
std = torch.zeros(3)
print('==> Computing mean and std..')
for inputs, targets in dataloader:
for i in range(3):
mean[i] += inputs[:,i,:,:].mean()
std[i] += inputs[:,i,:,:].std()
mean.div_(len(dataset))
std.div_(len(dataset))
return mean, std
def init_params(net):
'''Init layer parameters.'''
for m in net.modules():
if isinstance(m, nn.Conv2d):
init.kaiming_normal(m.weight, mode='fan_out')
if m.bias:
init.constant(m.bias, 0)
elif isinstance(m, nn.BatchNorm2d):
init.constant(m.weight, 1)
init.constant(m.bias, 0)
elif isinstance(m, nn.Linear):
init.normal(m.weight, std=1e-3)
if m.bias:
init.constant(m.bias, 0)
_, term_width = os.popen('stty size', 'r').read().split()
term_width = int(term_width)
TOTAL_BAR_LENGTH = 65.
last_time = time.time()
begin_time = last_time
def progress_bar(current, total, msg=None):
global last_time, begin_time
if current == 0:
begin_time = time.time() # Reset for new bar.
cur_len = int(TOTAL_BAR_LENGTH*current/total)
rest_len = int(TOTAL_BAR_LENGTH - cur_len) - 1
sys.stdout.write(' [')
for i in range(cur_len):
sys.stdout.write('=')
sys.stdout.write('>')
for i in range(rest_len):
sys.stdout.write('.')
sys.stdout.write(']')
cur_time = time.time()
step_time = cur_time - last_time
last_time = cur_time
tot_time = cur_time - begin_time
L = []
L.append(' Step: %s' % format_time(step_time))
L.append(' | Tot: %s' % format_time(tot_time))
if msg:
L.append(' | ' + msg)
msg = ''.join(L)
sys.stdout.write(msg)
for i in range(term_width-int(TOTAL_BAR_LENGTH)-len(msg)-3):
sys.stdout.write(' ')
# Go back to the center of the bar.
for i in range(term_width-int(TOTAL_BAR_LENGTH/2)+2):
sys.stdout.write('\b')
sys.stdout.write(' %d/%d ' % (current+1, total))
if current < total-1:
sys.stdout.write('\r')
else:
sys.stdout.write('\n')
sys.stdout.flush()
def format_time(seconds):
days = int(seconds / 3600/24)
seconds = seconds - days*3600*24
hours = int(seconds / 3600)
seconds = seconds - hours*3600
minutes = int(seconds / 60)
seconds = seconds - minutes*60
secondsf = int(seconds)
seconds = seconds - secondsf
millis = int(seconds*1000)
f = ''
i = 1
if days > 0:
f += str(days) + 'D'
i += 1
if hours > 0 and i <= 2:
f += str(hours) + 'h'
i += 1
if minutes > 0 and i <= 2:
f += str(minutes) + 'm'
i += 1
if secondsf > 0 and i <= 2:
f += str(secondsf) + 's'
i += 1
if millis > 0 and i <= 2:
f += str(millis) + 'ms'
i += 1
if f == '':
f = '0ms'
return f