-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathdsprites.py
192 lines (155 loc) · 7.22 KB
/
dsprites.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
import os
import numpy as np
from sklearn.utils.extmath import cartesian
from torch.utils.data import DataLoader, Dataset
from urllib import request
import torch
class SequenceDataset(Dataset):
def __init__(self, seq_len=None, **kwargs):
super().__init__(**kwargs)
self.seq_len = seq_len
self.seq_transform_idxs = None # set in child class
self.index_manager = None # set in child class
self.factor_sizes = None # set in child class
self.data = None # set in child class
def __len__(self):
return len(self.data)
def __getitem__(self, id):
index = torch.randint(len(self.seq_transform_idxs), (1,))
transform_idx = self.seq_transform_idxs[index]
img_seq = []
feat_seq = []
for idx in id:
x0 = self.data[idx]
x0_feat = self.index_manager.index_to_features(idx)
xt_feat = x0_feat.copy()
xt_feat1 = x0_feat.copy()
img_x_seq = [x0]
feat_x_seq = [x0_feat]
step = torch.randint(0,self.seq_len-1, (1,))
#for t in range(self.seq_len):
#print(step % self.factor_sizes[transform_idx])
xt_feat[transform_idx] = (x0_feat[transform_idx] + (step.cpu().numpy())) % self.factor_sizes[transform_idx]
xt_idx = self.index_manager.features_to_index(xt_feat)
xt = self.data[xt_idx]
img_x_seq.append(xt.copy())
feat_x_seq.append(xt_feat.copy())
xt_feat1[transform_idx] = (x0_feat[transform_idx] + (step.cpu().numpy()+1)) % self.factor_sizes[transform_idx]
xt_idx1 = self.index_manager.features_to_index(xt_feat1)
xt1 = self.data[xt_idx1]
img_x_seq.append(xt1.copy())
feat_x_seq.append(xt_feat1.copy())
img_seq.append(img_x_seq)
feat_seq.append(feat_x_seq)
img_seq = torch.tensor(img_seq).unsqueeze(1)
feat_seq = torch.tensor(feat_seq)
return img_seq, feat_seq, index, step
class IndexManger(object):
"""Index mapping from features to positions of state space atoms."""
def __init__(self, factor_sizes):
"""Index to latent (= features) space and vice versa.
Args:
factor_sizes: List of integers with the number of distinct values for each
of the factors.
"""
self.factor_sizes = np.array(factor_sizes)
self.num_total = np.prod(self.factor_sizes)
self.factor_bases = self.num_total / np.cumprod(self.factor_sizes)
self.index_to_feat = cartesian([np.array(list(range(i))) for i in self.factor_sizes])
def features_to_index(self, features):
"""Returns the indices in the input space for given factor configurations.
Args:
features: Numpy matrix where each row contains a different factor
configuration for which the indices in the input space should be
returned.
"""
#assert np.all((0 <= features) & (features <= self.factor_sizes))
index = np.array(np.dot(features, self.factor_bases), dtype=np.int64)
#assert np.all((0 <= index) & (index < self.num_total))
return index
def index_to_features(self, index):
#assert np.all((0 <= index) & (index < self.num_total))
features = self.index_to_feat[index]
#assert np.all((0 <= features) & (features <= self.factor_sizes))
return features
class DSpritesDataset(SequenceDataset):
"""
A PyTorch wrapper for the dSprites dataset by
Matthey et al. 2017. The dataset provides a 2D scene
with a sprite under different transformations:
# dim, type, #values avail.-range
* 0, color | 1 | 1-1
* 1, shape | 3 | 1-3
* 2, scale | 6 | 0.5-1.
* 3, orientation | 40 | 0-2pi
* 4, x-position | 32 | 0-1
* 5, y-position | 32 | 0-1
for details see https://github.com/deepmind/dsprites-dataset
"""
def __init__(self, dir='/home/akeller/repo/TECA/tvae/datasets/dsprites/',
seq_transforms=['orientation'],
avail_transforms=None,
max_transform_len=18,
**kwargs):
super().__init__(**kwargs)
self.url = "https://github.com/deepmind/dsprites-dataset/raw/master/dsprites_ndarray_co1sh3sc6or40x32y32_64x64.npz"
self.path = dir
self.path_npz = os.path.join(self.path, 'dsprites_ndarray_co1sh3sc6or40x32y32_64x64.npz')
print("Loading DSprites...")
try:
full_data = np.load(self.path_npz, encoding="latin1", allow_pickle=True)
except FileNotFoundError:
os.makedirs(self.path, exist_ok=True)
print(f'downloading dataset ... saving to {self.path_npz}')
request.urlretrieve(self.url, self.path_npz)
full_data = np.load(self.path_npz, encoding="latin1", allow_pickle=True)
self.data = full_data['imgs'].squeeze().astype(np.float32)
self.latents_values = full_data['latents_values']
self.latents_classes = full_data['latents_classes']
self.metadata = full_data['metadata'][()]
original_factor_sizes = [3, 6, 40, 32, 32]
speeds = [1, 1, 4, 4, 4]
start_idx = [0, 1, 0, 0, 0]
data_by_factor = self.data.reshape((*original_factor_sizes,) + (64, 64))
# Subselect Avail Transforms
if avail_transforms is None:
avail_transforms = seq_transforms
print(f"Removing all transforms but {avail_transforms}")
avail_transform_idxs = [self.metadata['latents_names'][1:].index(x) for x in avail_transforms]
data_indexer = [-1 if i not in avail_transform_idxs else np.s_[start_idx[i]:max_transform_len:speeds[i]] for i
in range(len(original_factor_sizes))]
self.data = data_by_factor[data_indexer]
self.factor_sizes = [1 if i not in avail_transform_idxs else self.data.shape[i] for i in
range(len(original_factor_sizes))]
self.data = self.data.reshape(-1, 64, 64)
# Ignore color index
self.seq_transforms = seq_transforms
self.seq_transform_idxs = [self.metadata['latents_names'][1:].index(x) for x in self.seq_transforms]
self.index_manager = IndexManger(self.factor_sizes)
print("Done")
def __len__(self):
return len(self.data)
def get_dataloader(dir='./',
seq_transforms=['posX', 'posY'],
avail_transforms=None,
seq_len=9,
max_transform_len=18,
batch_size=20):
train_data = DSpritesDataset(dir=dir, seq_transforms=seq_transforms,
avail_transforms=avail_transforms,
max_transform_len=max_transform_len,
seq_len=seq_len)
sampler = torch.utils.data.sampler.BatchSampler(
torch.utils.data.sampler.RandomSampler(train_data),
batch_size=batch_size,
drop_last=True)
return DataLoader(
train_data,
sampler=sampler,
generator = torch.Generator(device='cuda'),
)
if __name__ == "__main__":
data_loader = get_dataloader()
for x, f in data_loader:
print(x, f)
break