-
Notifications
You must be signed in to change notification settings - Fork 4
/
Copy patharguments.py
190 lines (161 loc) · 8.37 KB
/
arguments.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
# -*- coding: utf-8 -*-
import argparse
from sqlite3 import NotSupportedError
MODEL_LIST = [
"resnet50-1k",
"vit-b-1k",
"vit-b-22k",
"swin-b-1k",
"swin-b-22k",
"moco-v3-b-1k",
"clip-resnet50",
"clip-vit-b"
]
META_TRAIN_DATASETS = [
"sun397",
"stl10",
"fru92",
"veg200",
"oxford-iiit-pets",
"eurosat"
]
TASK_ADAPT_DATASETS = [
# original setting
"cifar10",
"cifar100",
"cub200",
"nabirds",
"oxford-flowers",
"stanford-dogs",
"stanford-cars",
"fgvc-aircraft",
"food101",
"dtd",
"svhn",
"gtsrb",
# vtab benchmark
"vtab-caltech101",
"vtab-cifar(num_classes=100)",
"vtab-dtd",
"vtab-oxford_flowers102",
"vtab-oxford_iiit_pet",
"vtab-patch_camelyon",
"vtab-sun397",
"vtab-svhn",
"vtab-resisc45",
"vtab-eurosat",
"vtab-dmlab",
"vtab-kitti(task=\"closest_vehicle_distance\")",
"vtab-smallnorb(predicted_attribute=\"label_azimuth\")",
"vtab-smallnorb(predicted_attribute=\"label_elevation\")",
"vtab-dsprites(predicted_attribute=\"label_x_position\",num_classes=16)",
"vtab-dsprites(predicted_attribute=\"label_orientation\",num_classes=16)",
"vtab-clevr(task=\"closest_object_distance\")",
"vtab-clevr(task=\"count_all\")",
"vtab-diabetic_retinopathy(config=\"btgraham-300\")"
]
DATASET_DIVERSITIES = {
"cifar10": 70.2,
"cifar100": 70.9,
"cub200": 76,
"nabirds": 74.8,
"oxford-flowers": 72.7,
"stanford-dogs": 73.4,
"stanford-cars": 70.5,
"fgvc-aircraft": 65.9,
"food101": 72.7,
"dtd": 78.7,
"svhn": 61.8,
"gtsrb": 67.5,
"sun397": 76.9,
"stl10": 74.1,
"fru92": 74.1,
"veg200": 71.5,
"oxford-iiit-pets": 72.4,
"eurosat": 64.6,
# vtab benchmark
"vtab-caltech101": None,
"vtab-cifar(num_classes=100)": None,
"vtab-dtd": None,
"vtab-oxford_flowers102": None,
"vtab-oxford_iiit_pet": None,
"vtab-patch_camelyon": None,
"vtab-sun397": None,
"vtab-svhn": None,
"vtab-resisc45": None,
"vtab-eurosat": None,
"vtab-dmlab": None,
"vtab-kitti(task=\"closest_vehicle_distance\")": None,
"vtab-smallnorb(predicted_attribute=\"label_azimuth\")": None,
"vtab-smallnorb(predicted_attribute=\"label_elevation\")": None,
"vtab-dsprites(predicted_attribute=\"label_x_position\",num_classes=16)": None,
"vtab-dsprites(predicted_attribute=\"label_orientation\",num_classes=16)": None,
"vtab-clevr(task=\"closest_object_distance\")": None,
"vtab-clevr(task=\"count_all\")": None,
"vtab-diabetic_retinopathy(config=\"btgraham-300\")": None
}
class Arguments:
def __init__(self, stage='task_adapting'):
self._parser = argparse.ArgumentParser(description='Diversity-Aware Meta Visual Prompting.')
self.add_common_args()
if stage == 'meta_training':
self.add_meta_train_args()
elif stage == "task_adapting":
self.add_task_adapt_args()
else:
raise NotSupportedError
def add_common_args(self):
### log related
self._parser.add_argument('--output_dir', type=str, default='')
### data related
self._parser.add_argument('--batch_size', type=int, default=128, help='Batch size in training')
self._parser.add_argument('--base_dir', type=str, default='/data-x/g12/huangqidong/')
self._parser.add_argument('--dataset_perc', default=1.0, type=float, help='Dataset percentage for usage [default: 1.0].')
self._parser.add_argument('--crop_size', default=224, type=int, help='Input size of images [default: 224].')
self._parser.add_argument('--diversities', type=dict, default=DATASET_DIVERSITIES, help='Diversity values of datasets.')
### prompt related
self._parser.add_argument('--pretrained_model', type=str, default='vit-b-1k', choices=MODEL_LIST)
self._parser.add_argument('--prompt_method', type=str, default='padding', choices=['padding', 'fixed_patch', 'random_patch'])
self._parser.add_argument('--prompt_size', type=int, default=30, help='Padding size for visual prompts.')
self._parser.add_argument('--wo_da', action='store_true', default=False, help='Without diversity-aware strategy [default: False].')
### model related
self._parser.add_argument('--drop', type=float, default=0.0, metavar='PCT', help='Dropout rate (default: 0.)')
self._parser.add_argument('--attn_drop_rate', type=float, default=0.0, metavar='PCT', help='Attention dropout rate (default: 0.)')
self._parser.add_argument('--drop_path', type=float, default=0.1, metavar='PCT', help='Drop path rate (default: 0.1)')
### others
self._parser.add_argument('--seed', type=int, default=2023, metavar='S', help='Random seed (default: 2023)')
self._parser.add_argument('--gpu_ids', type=int, default=0, help='Ids of GPUs to use.')
self._parser.add_argument('--num_gpus', type=int, default=1, help='Num of GPUs to use.')
self._parser.add_argument('--num_workers', type=int, default=4, help='Worker nums of data loading.')
self._parser.add_argument('--pin_memory', action='store_true', help='Pin CPU memory in DataLoader for more efficient (sometimes) transfer to GPU.')
self._parser.set_defaults(pin_memory=True)
self._parser.add_argument('--distributed', action='store_true', default=False, help='Whether to use the distributed mode [default: False].')
def add_meta_train_args(self):
### data related
self._parser.add_argument('--meta_datasets', type=list, default=META_TRAIN_DATASETS, help='The datasets selected for meta training')
self._parser.add_argument('--test_dataset', type=str, default='oxford-flowers', choices=TASK_ADAPT_DATASETS, help='The dataset selected for evaluation')
self._parser.add_argument('--adapt_method', type=str, default='prompt_wo_head', choices=['prompt_wo_head', 'prompt_w_head'])
### meta training related
self._parser.add_argument('--epochs', type=int, default=200, help='Nums of training epochs.')
self._parser.add_argument('--optimizer', type=str, default='Adam', help='Optimizer for training [default: Adam]')
self._parser.add_argument('--num_tasks', type=int, default=6, help='Meta batch size, namely task num')
self._parser.add_argument('--meta_lr', type=float, default=1.0, help='Meta-level outer learning rate')
self._parser.add_argument('--update_lr', type=float, default=1.0, help='Task-level inner update learning rate')
self._parser.add_argument('--weight_decay', type=float, default=1e-4, help='Task-level inner update weight decay rate')
self._parser.add_argument('--update_step', type=int, default=8, help='Task-level inner update steps')
self._parser.add_argument('--update_step_test', type=int, default=20, help='Update steps for finetunning')
self._parser.add_argument('--meta_optim_choose', type=str, default="reptile", choices=['reptile'], help='Choice for using which meta learning method')
self._parser.add_argument('--meta_step_size', type=float, default=1.0, help='Task-level outer update step size')
def add_task_adapt_args(self):
### data related
self._parser.add_argument('--test_dataset', type=str, default='oxford-flowers', choices=TASK_ADAPT_DATASETS, help='The dataset selected for evaluation')
self._parser.add_argument('--adapt_method', type=str, default='prompt_wo_head', choices=['prompt_wo_head', 'prompt_w_head'])
self._parser.add_argument('--checkpoint_dir', type=str, default='')
### tuning related
self._parser.add_argument('--epochs', type=int, default=50, help='Nums of training epochs.')
self._parser.add_argument('--optimizer', type=str, default='Adam', help='Optimizer for training [default: Adam]')
self._parser.add_argument('--lr', type=float, default=1e+4, help='Task adapting learning rate')
self._parser.add_argument('--weight_decay', type=float, default=0, help='Task adapting weight decay rate')
self._parser.add_argument('--eval_only', action='store_true', default=False, help='Evaluate only [default: False].')
def parser(self):
return self._parser