1
|
|
|
""" |
2
|
|
|
Most of this code is from torchvision. |
3
|
|
|
I will remove all this once verbosity is reduced. |
4
|
|
|
More info: https://github.com/pytorch/vision/issues/2830 |
5
|
|
|
""" |
6
|
|
|
|
7
|
|
|
import os |
8
|
|
|
import gzip |
9
|
|
|
import urllib |
10
|
|
|
import zipfile |
11
|
|
|
import hashlib |
12
|
|
|
import tarfile |
13
|
|
|
from typing import Optional |
14
|
|
|
|
15
|
|
|
from torch.hub import tqdm |
16
|
|
|
|
17
|
|
|
from .typing import TypePath |
18
|
|
|
|
19
|
|
|
|
20
|
|
|
def calculate_md5(fpath, chunk_size=1024 * 1024): |
21
|
|
|
md5 = hashlib.md5() |
22
|
|
|
with open(fpath, 'rb') as f: |
23
|
|
|
for chunk in iter(lambda: f.read(chunk_size), b''): |
24
|
|
|
md5.update(chunk) |
25
|
|
|
return md5.hexdigest() |
26
|
|
|
|
27
|
|
|
|
28
|
|
|
def check_md5(fpath, md5, **kwargs): |
29
|
|
|
return md5 == calculate_md5(fpath, **kwargs) |
30
|
|
|
|
31
|
|
|
|
32
|
|
|
def check_integrity(fpath, md5=None): |
33
|
|
|
if not os.path.isfile(fpath): |
34
|
|
|
return False |
35
|
|
|
if md5 is None: |
36
|
|
|
return True |
37
|
|
|
return check_md5(fpath, md5) |
38
|
|
|
|
39
|
|
|
|
40
|
|
|
def gen_bar_updater(): |
41
|
|
|
pbar = tqdm(total=None) |
42
|
|
|
|
43
|
|
|
def bar_update(count, block_size, total_size): |
44
|
|
|
if pbar.total is None and total_size: |
45
|
|
|
pbar.total = total_size |
46
|
|
|
progress_bytes = count * block_size |
47
|
|
|
pbar.update(progress_bytes - pbar.n) |
48
|
|
|
|
49
|
|
|
return bar_update |
50
|
|
|
|
51
|
|
|
|
52
|
|
|
# Adapted from torchvision, removing print statements |
53
|
|
|
def download_and_extract_archive( |
54
|
|
|
url: str, |
55
|
|
|
download_root: TypePath, |
56
|
|
|
extract_root: Optional[TypePath] = None, |
57
|
|
|
filename: Optional[TypePath] = None, |
58
|
|
|
md5: str = None, |
59
|
|
|
remove_finished: bool = False, |
60
|
|
|
) -> None: |
61
|
|
|
download_root = os.path.expanduser(download_root) |
62
|
|
|
if extract_root is None: |
63
|
|
|
extract_root = download_root |
64
|
|
|
if not filename: |
65
|
|
|
filename = os.path.basename(url) |
66
|
|
|
download_url(url, download_root, filename, md5) |
67
|
|
|
archive = os.path.join(download_root, filename) |
68
|
|
|
extract_archive(archive, extract_root, remove_finished) |
69
|
|
|
|
70
|
|
|
|
71
|
|
|
def _is_tarxz(filename): |
72
|
|
|
return filename.endswith('.tar.xz') |
73
|
|
|
|
74
|
|
|
|
75
|
|
|
def _is_tar(filename): |
76
|
|
|
return filename.endswith('.tar') |
77
|
|
|
|
78
|
|
|
|
79
|
|
|
def _is_targz(filename): |
80
|
|
|
return filename.endswith('.tar.gz') |
81
|
|
|
|
82
|
|
|
|
83
|
|
|
def _is_tgz(filename): |
84
|
|
|
return filename.endswith('.tgz') |
85
|
|
|
|
86
|
|
|
|
87
|
|
|
def _is_gzip(filename): |
88
|
|
|
return filename.endswith('.gz') and not filename.endswith('.tar.gz') |
89
|
|
|
|
90
|
|
|
|
91
|
|
|
def _is_zip(filename): |
92
|
|
|
return filename.endswith('.zip') |
93
|
|
|
|
94
|
|
|
|
95
|
|
|
def extract_archive(from_path, to_path=None, remove_finished=False): |
96
|
|
|
if to_path is None: |
97
|
|
|
to_path = os.path.dirname(from_path) |
98
|
|
|
|
99
|
|
|
if _is_tar(from_path): |
100
|
|
|
with tarfile.open(from_path, 'r') as tar: |
101
|
|
|
tar.extractall(path=to_path) |
102
|
|
|
elif _is_targz(from_path) or _is_tgz(from_path): |
103
|
|
|
with tarfile.open(from_path, 'r:gz') as tar: |
104
|
|
|
tar.extractall(path=to_path) |
105
|
|
|
elif _is_tarxz(from_path): |
106
|
|
|
with tarfile.open(from_path, 'r:xz') as tar: |
107
|
|
|
tar.extractall(path=to_path) |
108
|
|
|
elif _is_gzip(from_path): |
109
|
|
|
stem = os.path.splitext(os.path.basename(from_path))[0] |
110
|
|
|
to_path = os.path.join(to_path, stem) |
111
|
|
|
with open(to_path, 'wb') as out_f, gzip.GzipFile(from_path) as zip_f: |
112
|
|
|
out_f.write(zip_f.read()) |
113
|
|
|
elif _is_zip(from_path): |
114
|
|
|
with zipfile.ZipFile(from_path, 'r') as z: |
115
|
|
|
z.extractall(to_path) |
116
|
|
|
else: |
117
|
|
|
raise ValueError(f'Extraction of {from_path} not supported') |
118
|
|
|
|
119
|
|
|
if remove_finished: |
120
|
|
|
os.remove(from_path) |
121
|
|
|
|
122
|
|
|
|
123
|
|
|
# Adapted from torchvision, removing print statements |
124
|
|
|
def download_url( |
125
|
|
|
url: str, |
126
|
|
|
root: TypePath, |
127
|
|
|
filename: Optional[TypePath] = None, |
128
|
|
|
md5: str = None, |
129
|
|
|
) -> None: |
130
|
|
|
"""Download a file from a url and place it in root. |
131
|
|
|
|
132
|
|
|
Args: |
133
|
|
|
url: URL to download file from |
134
|
|
|
root: Directory to place downloaded file in |
135
|
|
|
filename: Name to save the file under. |
136
|
|
|
If ``None``, use the basename of the URL |
137
|
|
|
md5: MD5 checksum of the download. If None, do not check |
138
|
|
|
""" |
139
|
|
|
|
140
|
|
|
root = os.path.expanduser(root) |
141
|
|
|
if not filename: |
142
|
|
|
filename = os.path.basename(url) |
143
|
|
|
fpath = os.path.join(root, filename) |
144
|
|
|
os.makedirs(root, exist_ok=True) |
145
|
|
|
# check if file is already present locally |
146
|
|
|
if not check_integrity(fpath, md5): |
147
|
|
|
try: |
148
|
|
|
print('Downloading ' + url + ' to ' + fpath) # noqa: T001 |
149
|
|
|
urllib.request.urlretrieve( |
150
|
|
|
url, fpath, |
151
|
|
|
reporthook=gen_bar_updater() |
152
|
|
|
) |
153
|
|
|
except (urllib.error.URLError, OSError) as e: |
154
|
|
|
if url[:5] == 'https': |
155
|
|
|
url = url.replace('https:', 'http:') |
156
|
|
|
message = ( |
157
|
|
|
'Failed download. Trying https -> http instead.' |
158
|
|
|
' Downloading ' + url + ' to ' + fpath |
159
|
|
|
) |
160
|
|
|
print(message) # noqa: T001 |
161
|
|
|
urllib.request.urlretrieve( |
162
|
|
|
url, fpath, |
163
|
|
|
reporthook=gen_bar_updater() |
164
|
|
|
) |
165
|
|
|
else: |
166
|
|
|
raise e |
167
|
|
|
# check integrity of downloaded file |
168
|
|
|
if not check_integrity(fpath, md5): |
169
|
|
|
raise RuntimeError('File not found or corrupted.') |
170
|
|
|
|