-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathconfig.py
288 lines (236 loc) · 15.1 KB
/
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
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
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
# ------------------------------------------------------------------
"""
Main config file
Contact Person: Mohamad Hakam Shams Eddin <[email protected]>
Computer Vision Group - Institute of Computer Science III - University of Bonn
"""
# ------------------------------------------------------------------
import argparse
import pickle
import os
import datetime
# ------------------------------------------------------------------
def add_all_arguments(parser):
# --- general options ---
parser.add_argument('--seed', type=int, default=0, help='random seed')
parser.add_argument('--n_workers', type=int, default=8, help='number of workers for multiprocessing')
parser.add_argument('--pin_memory', type=bool, default=True,
help='allocate the loaded samples in GPU memory. Use it with training on GPU')
parser.add_argument('--batch_size', type=int, default=1, help='batch size')
parser.add_argument('--name', type=str, default='test', help='name of the experiment')
parser.add_argument('--dir_log', type=str, default=r'./log', help='log folder')
parser.add_argument('--root_CERRA', type=str, default=r'../CERRA',
help='root of the CERRA dataset')
parser.add_argument('--root_NOAA_CERRA', type=str, default=r'../NOAA_CERRA',
help='root of the NOAA CERRA dataset')
parser.add_argument('--root_ERA5_Land', type=str, default=r'../ERA5-Land',
help='root of the TerrSysMP_NET dataset')
parser.add_argument('--root_NOAA', type=str, default=r'../NOAA_CORDEX',
help='root of the NOAA ERA5-Land dataset')
parser.add_argument('--root_synthetic', type=str, default=r'../Synthetic/synthetic_CERRA',
help='root of the synthetic dataset')
parser.add_argument('--encoder', type=str, default='Mamba', help='name of the encoder model')
parser.add_argument('--classifier', type=str, default='CNN_3D', help='name of the classifier model')
parser.add_argument('--codebook', type=str, default='LFQ', help='name of the quantization layer')
parser.add_argument('--gpu_id', type=str, default="0, 1, 2, 3", help='gpu ids: i.e. 0 (0,1,2, use -1 for CPU)')
parser.add_argument('--nan_fill', type=float, default=0., help='a value to fill missing values')
# --- encoder ---
parser.add_argument('--in_channels_dynamic', type=int, default=6, help='number of input dynamic variables')
parser.add_argument('--in_channels', type=int, default=2, help='number of input channels per variable') # 2 for Era5 and CERRA, 1 for synthetic
parser.add_argument('--en_embed_dim', type=int, default=[16, 16], help='hidden dimensions in the encoder model')
parser.add_argument('--en_depths', type=int, default=[2, 1], help='number transformer blocks inside each layer')
parser.add_argument('--en_patch_size', type=int, default=(1, 1, 1),
help='patch size inside transformer. Keep it 1 for regression tasks')
parser.add_argument('--en_window_size', type=int, default=[(2, 4, 4), (8, 1, 1)], help='window size for self-attention/scanning')
parser.add_argument('--en_mlp_ratio', type=float, default=4., help='ratio of mlp hidden dim to embedding dim')
parser.add_argument('--en_drop_rate', type=float, default=0., help='dropout rate')
parser.add_argument('--en_drop_path_rate', type=float, default=0., help='stochastic depth rate')
parser.add_argument('--en_patch_norm', type=bool, default=False,
help='if True, add normalization after patch embedding')
parser.add_argument('--en_use_checkpoint', type=bool, default=False,
help='whether to use checkpointing to save memory')
# encoder Swin
parser.add_argument('--en_n_heads', type=int, default=[2, 2], help='number of heads for self-attention')
parser.add_argument('--en_attn_drop_rate', type=float, default=0.0, help='attention dropout rate')
parser.add_argument('--en_qkv_bias', type=bool, default=True,
help='if True, add a learnable bias to query, key, value')
parser.add_argument('--en_qk_scale', type=float, default=None,
help='override default qk scale of head_dim ** -0.5 if set')
# encoder Mamba
parser.add_argument('--d_state', type=int, default=[1, 1], help='SSM state expansion factor')
parser.add_argument('--d_conv', type=int, default=[3, 3], help='local convolution width')
parser.add_argument('--expand', type=int, default=[1, 1], help='d_inner expansion factor')
parser.add_argument('--dt_min', type=int, default=0.01, help='SSM dt_min')
parser.add_argument('--dt_max', type=int, default=0.1, help='SSM dt_max')
# --- vector quantization ---
parser.add_argument('--codebook_size', type=int, default=2, help='number of codes in the codebook')
parser.add_argument('--codebook_dim', type=int, default=16, help='input dimension for the codebook')
# --- classifier ---
parser.add_argument('--cls_dim', type=int, default=16, help='input dimension for the classifier')
parser.add_argument('--cls_drop_rate', type=float, default=0., help='drop rate for classification layer')
parser.add_argument('--en_de_pretrained', type=str, default=None,
help='pretrained model i.e. a trained model with best loss')
parser.add_argument('--years_train', type=str, default=[str(year) for year in range(1984, 2018)], help='years for training')
parser.add_argument('--years_val', type=str, default=['2018', '2019', '2020'], help='years for validation')
parser.add_argument('--years_test', type=str, default=['2021', '2022', '2023', '2024'], help='years for testing')
parser.add_argument('--times_train', type=tuple, default=(1, 52 * 34), help='time steps for training on synthetic data')
parser.add_argument('--times_val', type=tuple, default=(52 * 34 + 1, 52 * 40), help='time steps for validation on synthetic data')
parser.add_argument('--times_test', type=tuple, default=(52 * 40 + 1, 52 * 46), help='time steps for testing on synthetic data' )
parser.add_argument('--delta_t', type=int, default=8, help='number of weeks or time steps')
parser.add_argument('--window_size', type=int, default=1, help='scaling factor for resolution (2 means half)')
parser.add_argument('--threshold', type=float, default=26.0, help='VHI threshold')
parser.add_argument('--alpha', type=float, default=0.5, help='alpha to compute VHI')
parser.add_argument('--region', type=str, default='EUR-11', help='CORDEX region')
parser.add_argument('--x_min', type=int, default=0, help='start of grid extension in x direction') # None
parser.add_argument('--x_max', type=int, default=200, help='end of grid extension in x direction') # None
parser.add_argument('--y_min', type=int, default=0, help='start of grid extension in y direction') # None
parser.add_argument('--y_max', type=int, default=200, help='start of grid extension in y direction') # None
parser.add_argument('--is_shuffle', type=bool, default=False, help='if True, apply data shuffling')
parser.add_argument('--is_aug', type=bool, default=True, help='if True, apply data augmentation')
parser.add_argument('--is_norm', type=bool, default=True, help='if True, apply data normalization')
parser.add_argument('--is_clima_scale', type=bool, default=True, help='if True, apply data normalization with climatology')
parser.add_argument('--n_epochs', type=int, default=100, help='number of epochs for training')
parser.add_argument('--optimizer', type=str, default='Adam', help='optimizer for training')
parser.add_argument('--lr', type=float, default=1e-3, help='learning rate')
parser.add_argument('--weight_decay', type=float, default=0.003, help='weight decay')
parser.add_argument('--beta1', type=float, default=0.9, help='beta1 momentum term for Adam/AdamW')
parser.add_argument('--beta2', type=float, default=0.999, help='beta2 momentum term for Adam/AdamW')
parser.add_argument('--lr_scheduler', type=str, default='cosine', help='learning rate scheduler')
parser.add_argument('--lr_warmup', type=int, default=1e-6, help='learning rate for warmup')
parser.add_argument('--lr_warmup_epochs', type=int, default=2, help='number of epochs for warmup')
parser.add_argument('--lr_min', type=float, default=1e-5, help='minimum learning rate')
parser.add_argument('--lr_decay_step', type=int, default=20, help='learning rate step decay')
parser.add_argument('--lr_decay_rate', type=float, default=0.9, help='learning rate decay')
parser.add_argument('--lambda_ortho', type=float, default=10., help='orthogonality weight')
parser.add_argument('--lambda_commitment', type=float, default=3.0, help='commitment weight')
parser.add_argument('--lambda_anomaly', type=float, default=100.0, help='anomaly weight')
parser.add_argument('--lambda_entropy', type=float, default=0.1, help='entropy weight')
parser.add_argument('--diversity_gamma', type=float, default=0.1, help='diversity gamma weight')
# input variables
parser.add_argument('--variables', type=str,
default=[
'wdir10',
'si10',
'al',
'hcc',
'lcc',
'msl',
# 'mcc',
# 'skt',
# 'rsn',
# 'sde',
# 'sd',
# 'sp',
# 'sr',
# 'tcc',
# 'tciwv',
# 'r2',
# 't2m',
# 'liqvsm',
# 'sot',
# 'vsw',
# 'tp'
# 'var_01',
# 'var_02',
# 'var_03',
# 'var_04',
# 'var_05',
# 'var_06'
# 'd2m',
# 't2m',
# 'fal',
# 'sp',
# 'e',
# 'tp',
# 'skt',
# 'stl1',
# 'swvl1'
]
, help='input dynamic variables')
parser.add_argument('--variables_static', type=str,
default=[
# 'lsm',
# 'orog',
# 'dl',
# 'voltso',
# 'vwiltm',
# 'dis_water',
# 'slope',
'latitude',
'longitude'],
help='input static variables for CERRA')
return parser
def read_arguments(train=True, print=True, save=True):
parser = argparse.ArgumentParser()
parser = add_all_arguments(parser)
parser.add_argument('--phase', type=str, default='train')
config = parser.parse_args()
config.phase = 'train' if train else 'test'
if print:
print_options(config, parser)
if save:
save_options(config, parser)
return config
def save_options(config, parser):
if config.name is None or len(config.name) == 0:
config.name = str(datetime.datetime.now().strftime('%Y-%m-%d_%H-%M-%S'))
dir_log = os.path.join(config.dir_log, config.name)
os.makedirs(dir_log, exist_ok=True)
with open(dir_log + '/config.txt', 'wt') as config_file:
message = ''
message += '----------------- Options --------------- -------------------\n\n'
for k, v in sorted(vars(config).items()):
if k in ['variables', 'years_train', 'years_val', 'years_test', 'dir_log', 'root_CERRA', 'root_NOAA']:
continue
# comment = ''
default = parser.get_default(k)
# if v != default:
comment = '\t[default: %s]' % str(default)
message += '{:>25}: {:<20}{}\n'.format(str(k), str(v), comment)
comment = '\t[default: %s]' % str(parser.get_default('root_CERRA'))
message += '\n{:>25}: {:<20}{}\n'.format('root_CERRA', vars(config)['root_CERRA'], comment)
comment = '\t[default: %s]' % str(parser.get_default('root_NOAA'))
message += '\n{:>25}: {:<20}{}\n'.format('root_NOAA', vars(config)['root_NOAA'], comment)
comment = '\t[default: %s]' % str(parser.get_default('dir_log'))
message += '{:>25}: {:<20}{}\n'.format('dir_log', vars(config)['dir_log'], comment)
message += '\n----------------- Input Variables ------- -------------------'
message += '\n\n{}\n'.format(str(config.variables))
message += '\n----------------- Years ----------------- -------------------'
if config.phase == 'train':
message += '\n\nTraining: {}'.format(str(config.years_train))
message += '\nValidation: {}\n'.format(str(config.years_val))
else:
message += '\n\nTesting: {}\n'.format(str(config.years_test))
message += '\n----------------- End ------------------- -------------------'
config_file.write(message)
with open(dir_log + '/config.pkl', 'wb') as config_file:
pickle.dump(config, config_file)
def print_options(config, parser):
message = ''
message += '----------------- Options --------------- -------------------\n\n'
for k, v in sorted(vars(config).items()):
if k in ['variables', 'years_train', 'years_val', 'years_test', 'dir_log', 'root_CERRA', 'root_NOAA']:
continue
# comment = ''
default = parser.get_default(k)
# if v != default:
comment = '\t[default: %s]' % str(default)
message += '{:>25}: {:<20}{}\n'.format(str(k), str(v), comment)
comment = '\t[default: %s]' % str(parser.get_default('root_CERRA'))
message += '\n{:>25}: {:<20}{}\n'.format('root_CERRA', vars(config)['root_CERRA'], comment)
comment = '\t[default: %s]' % str(parser.get_default('root_NOAA'))
message += '\n{:>25}: {:<20}{}\n'.format('root_NOAA', vars(config)['root_NOAA'], comment)
comment = '\t[default: %s]' % str(parser.get_default('dir_log'))
message += '{:>25}: {:<20}{}\n'.format('dir_log', vars(config)['dir_log'], comment)
message += '\n----------------- Input Variables ------- -------------------'
message += '\n\n{}\n'.format(str(config.variables))
message += '\n----------------- Years ----------------- -------------------'
if config.phase == 'train':
message += '\n\nTraining: {}'.format(str(config.years_train))
message += '\nValidation: {}\n'.format(str(config.years_val))
else:
message += '\n\nTesting: {}\n'.format(str(config.years_test))
message += '\n----------------- End ------------------- -------------------'
print(message)
if __name__ == '__main__':
config = read_arguments(train=True, print=True, save=False)