-
Notifications
You must be signed in to change notification settings - Fork 22
/
__init__.py
172 lines (141 loc) · 6.1 KB
/
__init__.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
"""
LF-Font
Copyright (c) 2020-present NAVER Corp.
MIT license
"""
import torch
import random
from .lmdbutils import (load_lmdb, load_json, read_data_from_lmdb)
from .p2dataset import FactTrainDataset, FactTestDataset
from .p1dataset import CombTrainDataset, CombTestDataset, FixedRefDataset
from .datautils import cyclize, sample, uniform_sample
from torch.utils.data import DataLoader
def get_fact_trn_loader(env, env_get, cfg, train_dict, dec_dict, transform, **kwargs):
# avail_fonts = [os.path.splitext(name)[0] + ".ttf" for name in avail_fonts]
dset = FactTrainDataset(
env,
env_get,
train_dict,
dec_dict,
content_font=cfg.content_font,
n_comps=int(cfg.n_comps),
n_in_chars=int(cfg.n_in_chars),
n_each_chars=int(cfg.n_each_chars),
n_targets=int(cfg.n_targets),
transform=transform
)
if cfg.use_ddp:
sampler = torch.utils.data.distributed.DistributedSampler(dset)
kwargs["shuffle"] = False
else:
sampler = None
loader = DataLoader(dset, batch_size=cfg.batch_size, sampler=sampler,
collate_fn=dset.collate_fn, **kwargs)
return dset, loader
def get_fact_test_loader(env, env_get, target_dict, ref_unis, cfg, avails, dec_dict, transform, **kwargs):
dset = FactTestDataset(
env,
env_get,
target_dict,
ref_unis,
avails,
dec_dict,
content_font=cfg.content_font,
language=cfg.language,
transform=transform,
n_comps=int(cfg.n_comps),
n_shots=int(cfg.n_shots),
ret_targets=True,
)
loader = DataLoader(dset, batch_size=cfg.batch_size, collate_fn=dset.collate_fn, **kwargs)
return dset, loader
def get_comb_trn_loader(env, env_get, cfg, train_dict, dec_dict, transform, **kwargs):
# avail_fonts = [os.path.splitext(name)[0] + '.ttf' for name in avail_fonts]
dset = CombTrainDataset(
env,
env_get,
train_dict,
dec_dict,
content_font=cfg.content_font,
**cfg.get('dset_args', {}),
transform = transform,
)
if cfg.use_ddp:
sampler = torch.utils.data.distributed.DistributedSampler(dset)
kwargs["shuffle"] = False
kwargs["num_workers"] = 0
else:
sampler = None
loader = DataLoader(dset, batch_size=cfg.batch_size, sampler=sampler,
collate_fn=dset.collate_fn, **kwargs)
return dset, loader
def get_comb_test_loader(env, env_get, target_dict, cfg, avails, dec_dict, transform, **kwargs):
# avail_fonts = [os.path.splitext(name)[0] + '.ttf' for name in avail_fonts]
dset = CombTestDataset(
env,
env_get,
target_dict,
avails,
dec_dict,
content_font=cfg.content_font,
language=cfg.language,
transform=transform,
n_comps=int(cfg.n_comps),
ret_targets=True
)
loader = DataLoader(dset, batch_size=cfg.batch_size,
collate_fn=dset.collate_fn, **kwargs)
return dset, loader
def get_fixedref_loader(env, env_get, decompose, target_dict, ref_unis, rep_content, cfg, dec_dict, transform, **kwargs):
# avail_fonts = [os.path.splitext(name)[0] + '.ttf' for name in avail_fonts]
print([chr(int(uni, 16)) for uni in ref_unis])
dset = FixedRefDataset(env,
env_get,
target_dict,
ref_unis,
rep_content=rep_content,
decompose=decompose,
content_font=cfg.content_font,
language=cfg.language,
decompose_dict=dec_dict,
transform=transform,
ret_targets=True
)
loader = DataLoader(dset, batch_size=cfg.batch_size,
collate_fn=dset.collate_fn, **kwargs)
return dset, loader
def get_cv_comb_loaders(env, env_get, cfg, data_meta, dec_dict, transform, **kwargs):
n_unis = cfg.cv_n_unis
n_fonts = cfg.cv_n_fonts
ufs = uniform_sample(data_meta["test"]["unseen_fonts"], n_fonts)
sfs = uniform_sample(data_meta["test"]["seen_fonts"], n_fonts)
sus = uniform_sample(data_meta["test"]["seen_unis"], n_unis)
uus = uniform_sample(data_meta["test"]["unseen_unis"], n_unis)
sfuu_dict = {fname: uus for fname in sfs}
ufsu_dict = {fname: sus for fname in ufs}
ufuu_dict = {fname: uus for fname in ufs}
cv_loaders = {'sfuu': get_comb_test_loader(env, env_get, sfuu_dict, cfg, data_meta['valid'], dec_dict, transform, **kwargs)[1],
'ufsu': get_comb_test_loader(env, env_get, ufsu_dict, cfg, data_meta['valid'], dec_dict, transform, **kwargs)[1],
'ufuu': get_comb_test_loader(env, env_get, ufuu_dict, cfg, data_meta['valid'], dec_dict, transform, **kwargs)[1]
}
return cv_loaders
def get_cv_fact_loaders(env, env_get, cfg, data_meta, dec_dict, transform, ref_unis=None, **kwargs):
n_unis = cfg.cv_n_unis
n_fonts = cfg.cv_n_fonts
ufs = uniform_sample(data_meta["test"]["unseen_fonts"], n_fonts)
sfs = uniform_sample(data_meta["test"]["seen_fonts"], n_fonts)
sus = uniform_sample(data_meta["test"]["seen_unis"], n_unis)
uus = uniform_sample(data_meta["test"]["unseen_unis"], n_unis)
sfuu_dict = {fname: uus for fname in sfs}
ufsu_dict = {fname: sus for fname in ufs}
ufuu_dict = {fname: uus for fname in ufs}
if ref_unis is None:
ref_unis = sorted(set(data_meta["test"]["unseen_unis"]) - set(uus))[:cfg["n_shots"]]
cv_loaders = {"sfuu": get_fact_test_loader(env, env_get, sfuu_dict, ref_unis, cfg, data_meta["valid"],
dec_dict, transform, **kwargs)[1],
"ufsu": get_fact_test_loader(env, env_get, ufsu_dict, ref_unis, cfg, data_meta["valid"],
dec_dict, transform, **kwargs)[1],
"ufuu": get_fact_test_loader(env, env_get, ufuu_dict, ref_unis, cfg, data_meta["valid"],
dec_dict, transform, **kwargs)[1]
}
return cv_loaders