|
1
|
|
|
"""Most of this code is from torchvision. |
|
2
|
|
|
|
|
3
|
|
|
I will remove all this once verbosity is reduced. More info: |
|
4
|
|
|
https://github.com/pytorch/vision/issues/2830 |
|
5
|
|
|
""" |
|
6
|
|
|
|
|
7
|
|
|
from __future__ import annotations |
|
8
|
|
|
|
|
9
|
|
|
import gzip |
|
10
|
|
|
import hashlib |
|
11
|
|
|
import os |
|
12
|
|
|
import tarfile |
|
13
|
|
|
import urllib |
|
14
|
|
|
import zipfile |
|
15
|
|
|
|
|
16
|
|
|
from torch.hub import tqdm |
|
17
|
|
|
|
|
18
|
|
|
from .types import TypePath |
|
19
|
|
|
|
|
20
|
|
|
|
|
21
|
|
|
def calculate_md5(fpath, chunk_size=1024 * 1024): |
|
22
|
|
|
md5 = hashlib.md5() |
|
23
|
|
|
with open(fpath, 'rb') as f: |
|
24
|
|
|
for chunk in iter(lambda: f.read(chunk_size), b''): |
|
25
|
|
|
md5.update(chunk) |
|
26
|
|
|
return md5.hexdigest() |
|
27
|
|
|
|
|
28
|
|
|
|
|
29
|
|
|
def check_md5(fpath, md5, **kwargs): |
|
30
|
|
|
return md5 == calculate_md5(fpath, **kwargs) |
|
31
|
|
|
|
|
32
|
|
|
|
|
33
|
|
|
def check_integrity(fpath, md5=None): |
|
34
|
|
|
if not os.path.isfile(fpath): |
|
35
|
|
|
return False |
|
36
|
|
|
if md5 is None: |
|
37
|
|
|
return True |
|
38
|
|
|
return check_md5(fpath, md5) |
|
39
|
|
|
|
|
40
|
|
|
|
|
41
|
|
|
def gen_bar_updater(): |
|
42
|
|
|
pbar = tqdm(total=None) |
|
43
|
|
|
|
|
44
|
|
|
def bar_update(count, block_size, total_size): |
|
45
|
|
|
if pbar.total is None and total_size: |
|
46
|
|
|
pbar.total = total_size |
|
47
|
|
|
progress_bytes = count * block_size |
|
48
|
|
|
pbar.update(progress_bytes - pbar.n) |
|
49
|
|
|
|
|
50
|
|
|
return bar_update |
|
51
|
|
|
|
|
52
|
|
|
|
|
53
|
|
|
# Adapted from torchvision, removing print statements |
|
54
|
|
|
def download_and_extract_archive( |
|
55
|
|
|
url: str, |
|
56
|
|
|
download_root: TypePath, |
|
57
|
|
|
extract_root: TypePath | None = None, |
|
58
|
|
|
filename: TypePath | None = None, |
|
59
|
|
|
md5: str | None = None, |
|
60
|
|
|
remove_finished: bool = False, |
|
61
|
|
|
) -> None: |
|
62
|
|
|
download_root = os.path.expanduser(download_root) |
|
63
|
|
|
if extract_root is None: |
|
64
|
|
|
extract_root = download_root |
|
65
|
|
|
if not filename: |
|
66
|
|
|
filename = os.path.basename(url) |
|
67
|
|
|
download_url(url, download_root, filename, md5) |
|
68
|
|
|
archive = os.path.join(download_root, filename) |
|
69
|
|
|
extract_archive(archive, extract_root, remove_finished) |
|
70
|
|
|
|
|
71
|
|
|
|
|
72
|
|
|
def _is_tarxz(filename): |
|
73
|
|
|
return filename.endswith('.tar.xz') |
|
74
|
|
|
|
|
75
|
|
|
|
|
76
|
|
|
def _is_tar(filename): |
|
77
|
|
|
return filename.endswith('.tar') |
|
78
|
|
|
|
|
79
|
|
|
|
|
80
|
|
|
def _is_targz(filename): |
|
81
|
|
|
return filename.endswith('.tar.gz') |
|
82
|
|
|
|
|
83
|
|
|
|
|
84
|
|
|
def _is_tgz(filename): |
|
85
|
|
|
return filename.endswith('.tgz') |
|
86
|
|
|
|
|
87
|
|
|
|
|
88
|
|
|
def _is_gzip(filename): |
|
89
|
|
|
return filename.endswith('.gz') and not filename.endswith('.tar.gz') |
|
90
|
|
|
|
|
91
|
|
|
|
|
92
|
|
|
def _is_zip(filename): |
|
93
|
|
|
return filename.endswith('.zip') |
|
94
|
|
|
|
|
95
|
|
|
|
|
96
|
|
|
def extract_archive(from_path, to_path=None, remove_finished=False): |
|
97
|
|
|
if to_path is None: |
|
98
|
|
|
to_path = os.path.dirname(from_path) |
|
99
|
|
|
|
|
100
|
|
|
if _is_tar(from_path): |
|
101
|
|
|
with tarfile.open(from_path, 'r') as tar: |
|
102
|
|
|
tar.extractall(path=to_path) |
|
103
|
|
|
elif _is_targz(from_path) or _is_tgz(from_path): |
|
104
|
|
|
with tarfile.open(from_path, 'r:gz') as tar: |
|
105
|
|
|
tar.extractall(path=to_path) |
|
106
|
|
|
elif _is_tarxz(from_path): |
|
107
|
|
|
with tarfile.open(from_path, 'r:xz') as tar: |
|
108
|
|
|
tar.extractall(path=to_path) |
|
109
|
|
|
elif _is_gzip(from_path): |
|
110
|
|
|
stem = os.path.splitext(os.path.basename(from_path))[0] |
|
111
|
|
|
to_path = os.path.join(to_path, stem) |
|
112
|
|
|
with open(to_path, 'wb') as out_f, gzip.GzipFile(from_path) as zip_f: |
|
113
|
|
|
out_f.write(zip_f.read()) |
|
114
|
|
|
elif _is_zip(from_path): |
|
115
|
|
|
with zipfile.ZipFile(from_path, 'r') as z: |
|
116
|
|
|
z.extractall(to_path) |
|
117
|
|
|
else: |
|
118
|
|
|
raise ValueError(f'Extraction of {from_path} not supported') |
|
119
|
|
|
|
|
120
|
|
|
if remove_finished: |
|
121
|
|
|
os.remove(from_path) |
|
122
|
|
|
|
|
123
|
|
|
|
|
124
|
|
|
# Adapted from torchvision, removing print statements |
|
125
|
|
|
def download_url( |
|
126
|
|
|
url: str, |
|
127
|
|
|
root: TypePath, |
|
128
|
|
|
filename: TypePath | None = None, |
|
129
|
|
|
md5: str | None = None, |
|
130
|
|
|
) -> None: |
|
131
|
|
|
"""Download a file from a url and place it in root. |
|
132
|
|
|
|
|
133
|
|
|
Args: |
|
134
|
|
|
url: URL to download file from |
|
135
|
|
|
root: Directory to place downloaded file in |
|
136
|
|
|
filename: Name to save the file under. |
|
137
|
|
|
If ``None``, use the basename of the URL |
|
138
|
|
|
md5: MD5 checksum of the download. If None, do not check |
|
139
|
|
|
""" |
|
140
|
|
|
|
|
141
|
|
|
root = os.path.expanduser(root) |
|
142
|
|
|
if not filename: |
|
143
|
|
|
filename = os.path.basename(url) |
|
144
|
|
|
fpath = os.path.join(root, filename) |
|
145
|
|
|
os.makedirs(root, exist_ok=True) |
|
146
|
|
|
# check if file is already present locally |
|
147
|
|
|
if not check_integrity(fpath, md5): |
|
148
|
|
|
try: |
|
149
|
|
|
print('Downloading ' + url + ' to ' + fpath) # noqa: T201 |
|
150
|
|
|
urllib.request.urlretrieve( |
|
151
|
|
|
url, |
|
152
|
|
|
fpath, |
|
153
|
|
|
reporthook=gen_bar_updater(), |
|
154
|
|
|
) |
|
155
|
|
|
except (urllib.error.URLError, OSError) as e: |
|
156
|
|
|
if url[:5] == 'https': |
|
157
|
|
|
url = url.replace('https:', 'http:') |
|
158
|
|
|
message = ( |
|
159
|
|
|
'Failed download. Trying https -> http instead. Downloading ' |
|
160
|
|
|
+ url |
|
161
|
|
|
+ ' to ' |
|
162
|
|
|
+ fpath |
|
163
|
|
|
) |
|
164
|
|
|
print(message) # noqa: T201 |
|
165
|
|
|
urllib.request.urlretrieve( |
|
166
|
|
|
url, |
|
167
|
|
|
fpath, |
|
168
|
|
|
reporthook=gen_bar_updater(), |
|
169
|
|
|
) |
|
170
|
|
|
else: |
|
171
|
|
|
raise e |
|
172
|
|
|
# check integrity of downloaded file |
|
173
|
|
|
if not check_integrity(fpath, md5): |
|
174
|
|
|
raise RuntimeError('File not found or corrupted.') |
|
175
|
|
|
|
TorchIO-project /
torchio
