forked from DingXiaoH/ACNet
-
Notifications
You must be signed in to change notification settings - Fork 0
/
base_config.py
58 lines (52 loc) · 4.36 KB
/
base_config.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
from collections import namedtuple
from model_map import get_dataset_name_by_model_name
import numpy as np
BaseConfigByEpoch = namedtuple('BaseConfigByEpoch', ['network_type', 'dataset_name', 'dataset_subset', 'global_batch_size', 'num_node', 'device',
'weight_decay', 'weight_decay_bias', 'optimizer_type', 'momentum',
'bias_lr_factor', 'max_epochs', 'base_lr', 'lr_epoch_boundaries', 'lr_decay_factor', 'linear_final_lr', 'cosine_minimum',
'warmup_epochs', 'warmup_method', 'warmup_factor',
'ckpt_iter_period', 'tb_iter_period',
'output_dir', 'tb_dir',
'init_weights', 'save_weights',
'val_epoch_period', 'grad_accum_iters',
'deps',
'se_reduce_scale', 'se_layers'])
def get_baseconfig_by_epoch(network_type, dataset_name, dataset_subset, global_batch_size, num_node,
weight_decay, optimizer_type, momentum,
max_epochs, base_lr, lr_epoch_boundaries, lr_decay_factor, linear_final_lr, cosine_minimum,
warmup_epochs, warmup_method, warmup_factor,
ckpt_iter_period, tb_iter_period,
output_dir, tb_dir, save_weights,
device='cuda', weight_decay_bias=0, bias_lr_factor=2, init_weights=None, val_epoch_period=-1, grad_accum_iters=1,
deps=None,
se_reduce_scale=0, se_layers=None):
print('----------------- show lr schedule --------------')
print('base_lr:', base_lr)
print('max_epochs:', max_epochs)
print('lr_epochs:', lr_epoch_boundaries)
print('lr_decay:', lr_decay_factor)
print('linear_final_lr:', linear_final_lr)
print('-------------------------------------------------')
if deps is not None:
deps = np.array(deps, dtype=np.int)
return BaseConfigByEpoch(network_type=network_type,dataset_name=dataset_name,dataset_subset=dataset_subset,global_batch_size=global_batch_size,num_node=num_node, device=device,
weight_decay=weight_decay,weight_decay_bias=weight_decay_bias,optimizer_type=optimizer_type,momentum=momentum,bias_lr_factor=bias_lr_factor,
max_epochs=max_epochs, base_lr=base_lr, lr_epoch_boundaries=lr_epoch_boundaries,lr_decay_factor=lr_decay_factor, linear_final_lr=linear_final_lr, cosine_minimum=cosine_minimum,
warmup_epochs=warmup_epochs,warmup_method=warmup_method,warmup_factor=warmup_factor,
ckpt_iter_period=int(ckpt_iter_period),tb_iter_period=int(tb_iter_period),
output_dir=output_dir, tb_dir=tb_dir,
init_weights=init_weights, save_weights=save_weights,
val_epoch_period=val_epoch_period, grad_accum_iters=grad_accum_iters, deps=deps, se_reduce_scale=se_reduce_scale,
se_layers=se_layers)
def get_baseconfig_for_test(network_type, dataset_subset, global_batch_size, init_weights=None, device='cuda', deps=None,
se_reduce_scale=0, se_layers=None, dataset_name=None):
if dataset_name is None:
dataset_name = get_dataset_name_by_model_name(network_type)
return BaseConfigByEpoch(network_type=network_type, dataset_name=dataset_name,
dataset_subset=dataset_subset, global_batch_size=global_batch_size, num_node=1, device=device,
weight_decay=None, weight_decay_bias=None, optimizer_type=None, momentum=None, bias_lr_factor=None,
max_epochs=None, base_lr=None, lr_epoch_boundaries=None, lr_decay_factor=None, linear_final_lr=None, cosine_minimum=None,
warmup_epochs=None, warmup_method=None, warmup_factor=None, ckpt_iter_period=None,
tb_iter_period=None, output_dir=None, tb_dir=None, init_weights=init_weights,
save_weights=None, val_epoch_period=None, grad_accum_iters=None, deps=deps,
se_reduce_scale=se_reduce_scale, se_layers=se_layers)