-
Notifications
You must be signed in to change notification settings - Fork 5
/
options.py
181 lines (158 loc) · 9.36 KB
/
options.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
import os
import argparse
def get_training_config_and_args():
parser = get_training_parser()
args = parser.parse_args()
config = {}
for group in parser._action_groups:
if group.title in ['optional arguments', 'positional arguments']:
group_dict = {arg.dest: getattr(args, arg.dest, None) for arg in group._group_actions}
for key in group_dict:
config[key] = group_dict[key]
print(key)
else:
group_dict = {arg.dest: getattr(args, arg.dest, None) for arg in group._group_actions}
config[group.title] = group_dict
for dir in [config['callbacks']['log_dir'], config['callbacks']['save_dir']]:
if not os.path.exists(dir):
os.system(f"mkdir -p {dir}")
return config, args
def get_training_parser():
parser = argparse.ArgumentParser(description='Visual Dialog Toolkit',
formatter_class=argparse.ArgumentDefaultsHelpFormatter)
parser.add_argument('--seed', metavar='N', type=int,
default=0)
add_dataset_args(parser)
add_model_args(parser)
add_solver_args(parser)
add_callback_args(parser)
parser.add_argument('--config_name', metavar='S', default='c1.0.0')
return parser
def add_dataset_args(parser):
group = parser.add_argument_group("dataset")
group.add_argument('--v0.9', action='store_true',
default=False)
group.add_argument('--overfit', action='store_true',
default=False,
help='overfit on small dataset')
group.add_argument('--concat_hist', action='store_true',
default=False,
help='concat history rounds into a single vector')
group.add_argument('--max_seq_len', type=int, metavar='N',
default=20,
help='max number of tokens in a sequence')
group.add_argument('--vocab_min_count', type=int, metavar='N',
default=5,
help='The word with frequency of 5 times will be listed in Vocabulary')
group.add_argument('--finetune', default=False, action='store_true')
group.add_argument('--is_add_boundaries', default=True, action='store_true')
group.add_argument('--is_return_options', default=True, action='store_true')
group.add_argument('--num_boxes', choices=['fixed', 'adaptive'],
default='fixed', metavar='S',
help='The number of boxes per image from Faster R-CNN')
group.add_argument('--glove_path', metavar='PATH',
default='datasets/glove/embedding_Glove_840_300d.pkl')
group.add_argument('--train_feat_img_path', metavar='PATH',
default='datasets/bottom-up-attention/trainval_resnet101_faster_rcnn_genome_100.h5')
group.add_argument('--val_feat_img_path', metavar='PATH',
default='datasets/bottom-up-attention/val2018_resnet101_faster_rcnn_genome_100.h5')
group.add_argument('--test_feat_img_path', metavar='PATH',
default='datasets/bottom-up-attention/test2018_resnet101_faster_rcnn_genome_100.h5')
group.add_argument('--train_json_dialog_path', metavar='PATH',
default='datasets/annotations/visdial_1.0_train.json')
group.add_argument('--val_json_dialog_path', metavar='PATH',
default='datasets/annotations/visdial_1.0_val.json')
group.add_argument('--test_json_dialog_path', metavar='PATH',
default='datasets/annotations/visdial_1.0_test.json')
group.add_argument('--val_json_dense_dialog_path', metavar='PATH',
default='datasets/annotations/visdial_1.0_val_dense_annotations.json')
group.add_argument('--train_json_word_count_path', metavar='PATH',
default='datasets/annotations/visdial_1.0_word_counts_train.json')
return group
def add_solver_args(parser):
group = parser.add_argument_group('solver')
"""Adam Optimizer"""
group.add_argument('--optimizer', default='adam',
choices=['sgd', 'adam', 'adamax'])
group.add_argument('--adam_betas', nargs='+', type=float, default=[0.9, 0.997])
group.add_argument('--adam_eps', type=float, default=1e-9)
group.add_argument('--weight_decay', '--wd', default=1e-5, type=float, metavar='WD',
help='weight decay')
group.add_argument('--clip_norm', default=None, type=float,
metavar='N',
help='clip threshold of gradients')
"""Dataloader"""
group.add_argument('--num_epochs', default=30, type=int, metavar='N',
help='Total number of epochs')
group.add_argument('--batch_size', default=8, type=int,
metavar='N',
help="Batch_size for training")
group.add_argument('--cpu_workers', default=8, type=int)
group.add_argument('--batch_size_multiplier', default=1, type=int,
metavar='N',
help='Cumsum of loss in N batches and update optimizer once')
"""Learning Rate Scheduler"""
group.add_argument('--scheduler_type', default='LinearLR',
help='learning rate scheduler type',
choices=['CosineLR', 'LinearLR', "CosineStepLR"])
group.add_argument('--init_lr', default=5e-3, type=float,
help='initial learning rate')
group.add_argument('--min_lr', default=1e-5, type=float, metavar='LR',
help='minimum learning rate')
group.add_argument('--num_samples', default=123287, type=int,
help='The number of training samples')
"""Warmup Scheduler"""
group.add_argument('--warmup_factor', default=0.2, type=float,
metavar='N',
help='lr will increase from 0 -> init_lr with warm_factor:'
'after every batch, lr = lr * warmup_factor')
group.add_argument('--warmup_epochs', default=1, type=int, metavar='N')
"""Linear Scheduler"""
group.add_argument('--linear_gama', default=0.5, type=float, metavar='LG',
help='learning rate shrink factor for step reduce, lr_new = (lr * lr_gama) at milestone step')
group.add_argument('--milestone_steps', nargs='+', type=int, metavar='LS', default=[3, 6, 8, 10, 11],
help='If we use step_lr_scheduler rather than cosine')
group.add_argument('--fp16', default=False, action='store_true')
return group
def add_callback_args(parser):
group = parser.add_argument_group('callbacks')
group.add_argument('--resume', default=False, action='store_true')
group.add_argument('--validate', default=True, action='store_true')
group.add_argument('--path_pretrained_ckpt', metavar='DIR', default=None,
help='filename in save-dir from which to load checkpoint, checkpoint_last.pt')
group.add_argument('--save_dir', default='checkpoints/')
group.add_argument('--log_dir', default='checkpoints/tensorboard/')
return group
def add_model_args(parser):
group = parser.add_argument_group('model')
group.add_argument('--decoder_type', choices=['misc', 'disc', 'gen'], default='misc', help='Type of decoder')
group.add_argument('--encoder_out', type=str, nargs='+', default=['img', 'ques'], )
group.add_argument('--hidden_size', type=int, metavar='N', default=512)
group.add_argument('--dropout', type=float, metavar='N', default=0.1)
group.add_argument('--test_mode', action='store_true', default=False)
"""Image Feature"""
group.add_argument('--img_feat_size', type=int, metavar='N', default=2048)
group.add_argument('--img_num_attns', type=int, metavar='N', default=None)
group.add_argument('--img_has_bboxes', action='store_true', default=False)
group.add_argument('--img_has_attributes', action='store_true', default=False)
group.add_argument('--img_has_classes', action='store_true', default=False)
"""Text Feature"""
group.add_argument('--txt_vocab_size', type=int, metavar='N', default=11322)
group.add_argument('--txt_tokenizer', choices=['nlp', 'bert'], default='nlp')
group.add_argument('--txt_bidirectional', action='store_true', default=True)
group.add_argument('--txt_embedding_size', type=int, default=300)
group.add_argument('--txt_has_pos_embedding', action='store_true', default=False)
group.add_argument('--txt_has_layer_norm', action='store_true', default=False)
group.add_argument('--txt_has_decoder_layer_norm', action='store_true', default=False)
"""Cross-Attention"""
group.add_argument('--ca_has_shared_attns', action='store_true', default=False)
group.add_argument('--ca_has_proj_linear', action='store_true', default=False)
group.add_argument('--ca_has_layer_norm', action='store_true', default=False)
group.add_argument('--ca_has_residual', action='store_true', default=False)
group.add_argument('--ca_num_attn_stacks', type=int, metavar='N', default=1)
group.add_argument('--ca_num_attn_heads', type=int, metavar='N', default=4)
group.add_argument('--ca_pad_size', type=int, default=2)
# computing the avg attention maps for further visualization
group.add_argument('--ca_has_avg_attns', action='store_true', default=False)
group.add_argument('--ca_has_self_attns', action='store_true', default=False)
return group