mattdeitke
commited on
Commit
•
aae8281
1
Parent(s):
1e35c61
start refactor with abstract class
Browse files- objaverse_xl/abstract.py +82 -0
- objaverse_xl/github.py +486 -467
objaverse_xl/abstract.py
ADDED
@@ -0,0 +1,82 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from abc import ABC, abstractmethod
|
2 |
+
import pandas as pd
|
3 |
+
|
4 |
+
from typing import Optional, Callable
|
5 |
+
|
6 |
+
|
7 |
+
class ObjaverseSource(ABC):
|
8 |
+
@abstractmethod
|
9 |
+
def load_annotations(self, download_dir: str = "~/.objaverse") -> pd.DataFrame:
|
10 |
+
pass
|
11 |
+
|
12 |
+
@abstractmethod
|
13 |
+
def download_objects(
|
14 |
+
self,
|
15 |
+
objects: pd.DataFrame,
|
16 |
+
download_dir: str = "~/.objaverse",
|
17 |
+
processes: Optional[int] = None,
|
18 |
+
handle_found_object: Optional[Callable] = None,
|
19 |
+
handle_modified_object: Optional[Callable] = None,
|
20 |
+
handle_missing_object: Optional[Callable] = None,
|
21 |
+
**kwargs
|
22 |
+
) -> None:
|
23 |
+
"""Downloads all objects from the source.
|
24 |
+
|
25 |
+
Args:
|
26 |
+
objects (pd.DataFrmae): Objects to download. Must have columns for
|
27 |
+
the object "fileIdentifier" and "sha256". Use the `load_annotations`
|
28 |
+
function to get the metadata.
|
29 |
+
processes (Optional[int], optional): Number of processes to use for
|
30 |
+
downloading. If None, will use the number of CPUs on the machine.
|
31 |
+
Defaults to None.
|
32 |
+
download_dir (str, optional): Directory to download the objects to.
|
33 |
+
Supports all file systems supported by fsspec. Defaults to
|
34 |
+
"~/.objaverse".
|
35 |
+
save_repo_format (Optional[Literal["zip", "tar", "tar.gz", "files"]],
|
36 |
+
optional): Format to save the repository. If None, the repository will
|
37 |
+
not be saved. If "files" is specified, each file will be saved
|
38 |
+
individually. Otherwise, the repository can be saved as a "zip", "tar",
|
39 |
+
or "tar.gz" file. Defaults to None.
|
40 |
+
handle_found_object (Optional[Callable], optional): Called when an object is
|
41 |
+
successfully found and downloaded. Here, the object has the same sha256
|
42 |
+
as the one that was downloaded with Objaverse-XL. If None, the object
|
43 |
+
will be downloaded, but nothing will be done with it. Args for the
|
44 |
+
function include:
|
45 |
+
- local_path (str): Local path to the downloaded 3D object.
|
46 |
+
- file_identifier (str): File identifier of the 3D object.
|
47 |
+
- sha256 (str): SHA256 of the contents of the 3D object.
|
48 |
+
- metadata (Dict[Hashable, Any]): Metadata about the 3D object,
|
49 |
+
including the GitHub organization and repo names.
|
50 |
+
Return is not used. Defaults to None.
|
51 |
+
handle_modified_object (Optional[Callable], optional): Called when a
|
52 |
+
modified object is found and downloaded. Here, the object is
|
53 |
+
successfully downloaded, but it has a different sha256 than the one that
|
54 |
+
was downloaded with Objaverse-XL. This is not expected to happen very
|
55 |
+
often, because the same commit hash is used for each repo. If None, the
|
56 |
+
object will be downloaded, but nothing will be done with it. Args for
|
57 |
+
the function include:
|
58 |
+
- local_path (str): Local path to the downloaded 3D object.
|
59 |
+
- file_identifier (str): File identifier of the 3D object.
|
60 |
+
- new_sha256 (str): SHA256 of the contents of the newly downloaded 3D
|
61 |
+
object.
|
62 |
+
- old_sha256 (str): Expected SHA256 of the contents of the 3D object as
|
63 |
+
it was when it was downloaded with Objaverse-XL.
|
64 |
+
- metadata (Dict[Hashable, Any]): Metadata about the 3D object, which is
|
65 |
+
particular to the souce.
|
66 |
+
Return is not used. Defaults to None.
|
67 |
+
handle_missing_object (Optional[Callable], optional): Called when an object
|
68 |
+
that is in Objaverse-XL is not found. Here, it is likely that the
|
69 |
+
repository was deleted or renamed. If None, nothing will be done with
|
70 |
+
the missing object.
|
71 |
+
Args for the function include:
|
72 |
+
- file_identifier (str): File identifier of the 3D object.
|
73 |
+
- sha256 (str): SHA256 of the contents of the original 3D object.
|
74 |
+
- metadata (Dict[Hashable, Any]): Metadata about the 3D object, which is
|
75 |
+
particular to the source.
|
76 |
+
Return is not used. Defaults to None.
|
77 |
+
|
78 |
+
Returns:
|
79 |
+
List[Dict[str, str]]: List of dictionaries with the keys "fileIdentifier"
|
80 |
+
and "sha256" for each downloaded object.
|
81 |
+
"""
|
82 |
+
pass
|
objaverse_xl/github.py
CHANGED
@@ -10,6 +10,7 @@ import tempfile
|
|
10 |
from multiprocessing import Pool
|
11 |
from typing import Callable, Dict, List, Literal, Optional
|
12 |
from objaverse_xl.utils import get_file_hash
|
|
|
13 |
|
14 |
import fsspec
|
15 |
import pandas as pd
|
@@ -33,512 +34,530 @@ FILE_EXTENSIONS = [
|
|
33 |
]
|
34 |
|
35 |
|
36 |
-
|
37 |
-
"""
|
38 |
|
39 |
-
|
40 |
-
|
41 |
-
Supports all file systems supported by fsspec. Defaults to "~/.objaverse".
|
42 |
|
43 |
-
|
44 |
-
|
45 |
-
|
46 |
-
|
47 |
-
filename = os.path.join(download_dir, "github", "github-urls.parquet")
|
48 |
-
fs, path = fsspec.core.url_to_fs(filename)
|
49 |
-
fs.makedirs(os.path.dirname(path), exist_ok=True)
|
50 |
|
51 |
-
|
52 |
-
|
53 |
-
|
|
|
|
|
|
|
|
|
|
|
54 |
|
55 |
-
|
56 |
-
|
57 |
-
|
58 |
-
file.write(response.content)
|
59 |
|
60 |
-
|
61 |
-
|
62 |
-
|
|
|
63 |
|
64 |
-
|
|
|
|
|
65 |
|
66 |
-
|
67 |
|
|
|
68 |
|
69 |
-
def _get_repo_id_with_hash(item: pd.Series) -> str:
|
70 |
-
|
71 |
-
|
72 |
-
|
73 |
|
|
|
|
|
74 |
|
75 |
-
|
76 |
-
|
77 |
-
|
78 |
-
Args:
|
79 |
-
repo_url (str): URL of the repo to clone.
|
80 |
-
target_directory (str): Directory to clone the repo to.
|
81 |
-
|
82 |
-
Returns:
|
83 |
-
bool: True if the clone was successful, False otherwise.
|
84 |
-
"""
|
85 |
-
return _run_command_with_check(
|
86 |
-
["git", "clone", "--depth", "1", repo_url, target_directory],
|
87 |
-
)
|
88 |
|
|
|
|
|
|
|
|
|
|
|
|
|
89 |
|
90 |
-
def _run_command_with_check(
|
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 |
-
base_dir (str): Base directory to save the repo to.
|
135 |
-
expected_objects (Dict[str, str]): Dictionary of objects that one expects to
|
136 |
-
find in the repo. Keys are the "fileIdentifier" (i.e., the GitHub URL in
|
137 |
-
this case) and values are the "sha256" of the objects.
|
138 |
-
{and the rest of the args are the same as download_github_objects}
|
139 |
-
|
140 |
-
Returns:
|
141 |
-
List[Dict[str, str]]: List of dictionaries with the keys "fileIdentifier" and
|
142 |
-
"sha256" for each downloaded object.
|
143 |
-
"""
|
144 |
-
# NOTE: assuming that the user has already checked that the repo doesn't exist,
|
145 |
-
org, repo = repo_id.split("/")
|
146 |
-
|
147 |
-
with tempfile.TemporaryDirectory() as temp_dir:
|
148 |
-
# clone the repo to a temp directory
|
149 |
-
target_directory = os.path.join(temp_dir, repo)
|
150 |
-
successful_clone = _git_shallow_clone(
|
151 |
-
f"https://github.com/{org}/{repo}.git", target_directory
|
152 |
-
)
|
153 |
-
if not successful_clone:
|
154 |
-
logger.error(f"Could not clone {repo_id}")
|
155 |
-
if handle_missing_object is not None:
|
156 |
-
for github_url, sha256 in expected_objects.items():
|
157 |
handle_missing_object(
|
158 |
file_identifier=github_url,
|
159 |
sha256=sha256,
|
160 |
metadata=dict(github_organization=org, github_repo=repo),
|
161 |
)
|
162 |
-
return []
|
163 |
-
|
164 |
-
# use the commit hash if specified
|
165 |
-
repo_commit_hash = _get_commit_hash_from_local_git_dir(target_directory)
|
166 |
-
if commit_hash is not None:
|
167 |
-
keep_going = True
|
168 |
-
if repo_commit_hash != commit_hash:
|
169 |
-
# run git reset --hard && git checkout 37f4d8d287e201ce52c048bf74d46d6a09d26b2c
|
170 |
-
if not _run_command_with_check(
|
171 |
-
["git", "fetch", "origin", commit_hash], target_directory
|
172 |
-
):
|
173 |
-
logger.error(
|
174 |
-
f"Error in git fetch! Sticking with {repo_commit_hash=} instead of {commit_hash=}"
|
175 |
-
)
|
176 |
-
keep_going = False
|
177 |
|
178 |
-
|
179 |
-
["git", "reset", "--hard"], target_directory
|
180 |
-
):
|
181 |
-
logger.error(
|
182 |
-
f"Error in git reset! Sticking with {repo_commit_hash=} instead of {commit_hash=}"
|
183 |
-
)
|
184 |
-
keep_going = False
|
185 |
|
186 |
-
|
187 |
-
|
188 |
-
|
189 |
-
|
190 |
-
|
191 |
-
else:
|
192 |
-
logger.error(
|
193 |
-
f"Error in git checkout! Sticking with {repo_commit_hash=} instead of {commit_hash=}"
|
194 |
-
)
|
195 |
-
|
196 |
-
# pull the lfs files
|
197 |
-
_pull_lfs_files(target_directory)
|
198 |
-
|
199 |
-
# get all the files in the repo
|
200 |
-
files = _list_files(target_directory)
|
201 |
-
files_with_3d_extension = [
|
202 |
-
file
|
203 |
-
for file in files
|
204 |
-
if any(file.lower().endswith(ext) for ext in FILE_EXTENSIONS)
|
205 |
]
|
206 |
|
207 |
-
|
208 |
-
|
209 |
-
|
210 |
-
|
211 |
-
|
212 |
-
|
213 |
-
|
214 |
-
|
215 |
-
|
216 |
-
|
217 |
-
|
218 |
-
|
219 |
-
|
220 |
-
if expected_objects[github_url] == file_hash:
|
221 |
-
if handle_found_object is not None:
|
222 |
-
handle_found_object(
|
223 |
-
local_path=file,
|
224 |
-
file_identifier=github_url,
|
225 |
-
sha256=file_hash,
|
226 |
-
metadata=dict(github_organization=org, github_repo=repo),
|
227 |
-
)
|
228 |
-
else:
|
229 |
-
if handle_modified_object is not None:
|
230 |
-
handle_modified_object(
|
231 |
-
local_path=file,
|
232 |
-
file_identifier=github_url,
|
233 |
-
new_sha256=file_hash,
|
234 |
-
old_sha256=expected_objects[github_url],
|
235 |
-
metadata=dict(github_organization=org, github_repo=repo),
|
236 |
-
)
|
237 |
-
elif handle_new_object is not None:
|
238 |
-
handle_new_object(
|
239 |
-
local_path=file,
|
240 |
-
file_identifier=github_url,
|
241 |
-
sha256=file_hash,
|
242 |
-
metadata=dict(github_organization=org, github_repo=repo),
|
243 |
-
)
|
244 |
-
|
245 |
-
# save the file hashes to a json file
|
246 |
-
with open(
|
247 |
-
os.path.join(target_directory, ".objaverse-file-hashes.json"),
|
248 |
-
"w",
|
249 |
-
encoding="utf-8",
|
250 |
-
) as f:
|
251 |
-
json.dump(file_hashes, f, indent=2)
|
252 |
-
|
253 |
-
# remove the .git directory
|
254 |
-
shutil.rmtree(os.path.join(target_directory, ".git"))
|
255 |
-
|
256 |
-
if save_repo_format is not None:
|
257 |
-
logger.debug(f"Saving as {save_repo_format}")
|
258 |
-
# save the repo to a zip file
|
259 |
-
if save_repo_format == "zip":
|
260 |
-
shutil.make_archive(target_directory, "zip", target_directory)
|
261 |
-
elif save_repo_format == "tar":
|
262 |
-
with tarfile.open(os.path.join(temp_dir, f"{repo}.tar"), "w") as tar:
|
263 |
-
tar.add(target_directory, arcname=repo)
|
264 |
-
elif save_repo_format == "tar.gz":
|
265 |
-
with tarfile.open(
|
266 |
-
os.path.join(temp_dir, f"{repo}.tar.gz"), "w:gz"
|
267 |
-
) as tar:
|
268 |
-
tar.add(target_directory, arcname=repo)
|
269 |
-
elif save_repo_format == "files":
|
270 |
-
pass
|
271 |
-
else:
|
272 |
-
raise ValueError(
|
273 |
-
f"save_repo_format must be one of zip, tar, tar.gz, files. Got {save_repo_format}"
|
274 |
-
)
|
275 |
-
|
276 |
-
dirname = os.path.join(base_dir, "repos", org)
|
277 |
-
fs.makedirs(dirname, exist_ok=True)
|
278 |
-
if save_repo_format != "files":
|
279 |
-
# move the repo to the correct location (with put)
|
280 |
-
fs.put(
|
281 |
-
os.path.join(temp_dir, f"{repo}.{save_repo_format}"),
|
282 |
-
os.path.join(dirname, f"{repo}.{save_repo_format}"),
|
283 |
-
)
|
284 |
-
else:
|
285 |
-
# move the repo to the correct location (with put)
|
286 |
-
fs.put(target_directory, dirname, recursive=True)
|
287 |
-
|
288 |
-
# get each object that was missing from the expected objects
|
289 |
-
if handle_missing_object is not None:
|
290 |
-
obtained_urls = {x["fileIdentifier"] for x in file_hashes}
|
291 |
-
for github_url, sha256 in expected_objects.items():
|
292 |
-
if github_url not in obtained_urls:
|
293 |
-
handle_missing_object(
|
294 |
-
file_identifier=github_url,
|
295 |
-
sha256=sha256,
|
296 |
-
metadata=dict(github_organization=org, github_repo=repo),
|
297 |
-
)
|
298 |
-
|
299 |
-
return file_hashes
|
300 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
301 |
|
302 |
-
def
|
303 |
-
|
304 |
-
os.path.join(root, f) for root, dirs, files in os.walk(root_dir) for f in files
|
305 |
-
]
|
306 |
|
|
|
|
|
307 |
|
308 |
-
|
309 |
-
|
310 |
-
subprocess.run(["git", "lfs", "pull"], cwd=repo_dir, check=True)
|
311 |
|
|
|
|
|
|
|
|
|
312 |
|
313 |
-
def _has_lfs_files(repo_dir: str) -> bool:
|
314 |
-
gitattributes_path = os.path.join(repo_dir, ".gitattributes")
|
315 |
-
if not os.path.exists(gitattributes_path):
|
316 |
-
return False
|
317 |
-
with open(gitattributes_path, "r", encoding="utf-8") as f:
|
318 |
-
for line in f:
|
319 |
-
if "filter=lfs" in line:
|
320 |
-
return True
|
321 |
-
return False
|
322 |
-
|
323 |
-
|
324 |
-
def _get_commit_hash_from_local_git_dir(local_git_dir: str) -> str:
|
325 |
-
# get the git hash of the repo
|
326 |
-
result = subprocess.run(
|
327 |
-
["git", "rev-parse", "HEAD"], cwd=local_git_dir, capture_output=True, check=True
|
328 |
-
)
|
329 |
-
commit_hash = result.stdout.strip().decode("utf-8")
|
330 |
-
return commit_hash
|
331 |
-
|
332 |
-
|
333 |
-
def _parallel_process_repo(args) -> List[Dict[str, str]]:
|
334 |
-
"""Helper function to process a repo in parallel.
|
335 |
-
|
336 |
-
Note: This function is used to parallelize the processing of repos. It is not
|
337 |
-
intended to be called directly.
|
338 |
-
|
339 |
-
Args:
|
340 |
-
args (Tuple): Tuple of arguments to pass to _process_repo.
|
341 |
-
|
342 |
-
Returns:
|
343 |
-
List[Dict[str, str]]: List of dictionaries with the keys "fileIdentifier" and
|
344 |
-
"sha256" for each downloaded object.
|
345 |
-
"""
|
346 |
-
|
347 |
-
(
|
348 |
-
repo_id_hash,
|
349 |
-
fs,
|
350 |
-
base_dir,
|
351 |
-
save_repo_format,
|
352 |
-
expected_objects,
|
353 |
-
handle_found_object,
|
354 |
-
handle_modified_object,
|
355 |
-
handle_missing_object,
|
356 |
-
handle_new_object,
|
357 |
-
) = args
|
358 |
-
repo_id = "/".join(repo_id_hash.split("/")[:2])
|
359 |
-
commit_hash = repo_id_hash.split("/")[2]
|
360 |
-
return _process_repo(
|
361 |
-
repo_id=repo_id,
|
362 |
-
fs=fs,
|
363 |
-
base_dir=base_dir,
|
364 |
-
save_repo_format=save_repo_format,
|
365 |
-
expected_objects=expected_objects,
|
366 |
-
handle_found_object=handle_found_object,
|
367 |
-
handle_modified_object=handle_modified_object,
|
368 |
-
handle_missing_object=handle_missing_object,
|
369 |
-
handle_new_object=handle_new_object,
|
370 |
-
commit_hash=commit_hash,
|
371 |
-
)
|
372 |
-
|
373 |
-
|
374 |
-
def _process_group(group):
|
375 |
-
key, group_df = group
|
376 |
-
return key, group_df.set_index("fileIdentifier")["sha256"].to_dict()
|
377 |
-
|
378 |
-
|
379 |
-
def download_github_objects(
|
380 |
-
objects: pd.DataFrame,
|
381 |
-
processes: Optional[int] = None,
|
382 |
-
download_dir: str = "~/.objaverse",
|
383 |
-
save_repo_format: Optional[Literal["zip", "tar", "tar.gz", "files"]] = None,
|
384 |
-
handle_found_object: Optional[Callable] = None,
|
385 |
-
handle_modified_object: Optional[Callable] = None,
|
386 |
-
handle_missing_object: Optional[Callable] = None,
|
387 |
-
handle_new_object: Optional[Callable] = None,
|
388 |
-
) -> List[Dict[str, str]]:
|
389 |
-
"""Download the specified GitHub objects.
|
390 |
-
|
391 |
-
Args:
|
392 |
-
objects (pd.DataFrmae): GitHub objects to download. Must have columns for the
|
393 |
-
object "fileIdentifier" and "sha256". Use the load_github_metadata function
|
394 |
-
to get the metadata.
|
395 |
-
processes (Optional[int], optional): Number of processes to use for downloading.
|
396 |
-
If None, will use the number of CPUs on the machine. Defaults to None.
|
397 |
-
download_dir (str, optional): Directory to download the GitHub objects to.
|
398 |
-
Supports all file systems supported by fsspec. Defaults to "~/.objaverse".
|
399 |
-
save_repo_format (Optional[Literal["zip", "tar", "tar.gz", "files"]], optional):
|
400 |
-
Format to save the repository. If None, the repository will not be saved. If
|
401 |
-
"files" is specified, each file will be saved individually. Otherwise, the
|
402 |
-
repository can be saved as a "zip", "tar", or "tar.gz" file. Defaults to
|
403 |
-
None.
|
404 |
-
handle_found_object (Optional[Callable], optional): Called when an object is
|
405 |
-
successfully found and downloaded. Here, the object has the same sha256 as
|
406 |
-
the one that was downloaded with Objaverse-XL. If None, the object will be
|
407 |
-
downloaded, but nothing will be done with it. Args for the function include:
|
408 |
-
- local_path (str): Local path to the downloaded 3D object.
|
409 |
-
- file_identifier (str): GitHub URL of the 3D object.
|
410 |
-
- sha256 (str): SHA256 of the contents of the 3D object.
|
411 |
-
- metadata (Dict[str, Any]): Metadata about the 3D object, including the
|
412 |
-
GitHub organization and repo names.
|
413 |
-
Return is not used. Defaults to None.
|
414 |
-
handle_new_object (Optional[Callable], optional): Called when a new object is
|
415 |
-
found. Here, the object is not used in Objaverse-XL, but is still downloaded
|
416 |
-
with the repository. The object may have not been used because it does not
|
417 |
-
successfully import into Blender. If None, the object will be downloaded,
|
418 |
-
but nothing will be done with it. Args for the function include:
|
419 |
-
- local_path (str): Local path to the downloaded 3D object.
|
420 |
-
- file_identifier (str): GitHub URL of the 3D object.
|
421 |
-
- sha256 (str): SHA256 of the contents of the 3D object.
|
422 |
-
- metadata (Dict[str, Any]): Metadata about the 3D object, including the
|
423 |
-
GitHub organization and repo names.
|
424 |
-
Return is not used. Defaults to None.
|
425 |
-
handle_modified_object (Optional[Callable], optional): Called when a modified
|
426 |
-
object is found and downloaded. Here, the object is successfully downloaded,
|
427 |
-
but it has a different sha256 than the one that was downloaded with
|
428 |
-
Objaverse-XL. This is not expected to happen very often, because the same
|
429 |
-
commit hash is used for each repo. If None, the object will be downloaded,
|
430 |
-
but nothing will be done with it. Args for the function include:
|
431 |
-
- local_path (str): Local path to the downloaded 3D object.
|
432 |
-
- file_identifier (str): GitHub URL of the 3D object.
|
433 |
-
- new_sha256 (str): SHA256 of the contents of the newly downloaded 3D
|
434 |
-
object.
|
435 |
-
- old_sha256 (str): Expected SHA256 of the contents of the 3D object as it
|
436 |
-
was when it was downloaded with Objaverse-XL.
|
437 |
-
- metadata (Dict[str, Any]): Metadata about the 3D object, including the
|
438 |
-
GitHub organization and repo names.
|
439 |
-
Return is not used. Defaults to None.
|
440 |
-
handle_missing_object (Optional[Callable], optional): Called when an object that
|
441 |
-
is in Objaverse-XL is not found. Here, it is likely that the repository was
|
442 |
-
deleted or renamed. If None, nothing will be done with the missing object.
|
443 |
-
Args for the function include:
|
444 |
-
- file_identifier (str): GitHub URL of the 3D object.
|
445 |
-
- sha256 (str): SHA256 of the contents of the original 3D object.
|
446 |
-
- metadata (Dict[str, Any]): Metadata about the 3D object, including the
|
447 |
-
GitHub organization and repo names.
|
448 |
-
Return is not used. Defaults to None.
|
449 |
-
|
450 |
-
Returns:
|
451 |
-
List[Dict[str, str]]: List of dictionaries with the keys "fileIdentifier" and
|
452 |
-
"sha256" for each downloaded object.
|
453 |
-
"""
|
454 |
-
if processes is None:
|
455 |
-
processes = multiprocessing.cpu_count()
|
456 |
-
|
457 |
-
base_download_dir = os.path.join(download_dir, "github")
|
458 |
-
fs, path = fsspec.core.url_to_fs(base_download_dir)
|
459 |
-
fs.makedirs(path, exist_ok=True)
|
460 |
-
|
461 |
-
# Getting immediate subdirectories of root_path
|
462 |
-
if save_repo_format == "files":
|
463 |
-
downloaded_repo_dirs = fs.glob(base_download_dir + "/repos/*/*/")
|
464 |
-
downloaded_repo_ids = {
|
465 |
-
"/".join(x.split("/")[-2:]) for x in downloaded_repo_dirs
|
466 |
-
}
|
467 |
-
else:
|
468 |
-
downloaded_repo_dirs = fs.glob(
|
469 |
-
base_download_dir + f"/repos/*/*.{save_repo_format}"
|
470 |
-
)
|
471 |
-
downloaded_repo_ids = set()
|
472 |
-
for x in downloaded_repo_dirs:
|
473 |
-
org, repo = x.split("/")[-2:]
|
474 |
-
repo = repo[: -len(f".{save_repo_format}")]
|
475 |
-
repo_id = f"{org}/{repo}"
|
476 |
-
downloaded_repo_ids.add(repo_id)
|
477 |
-
|
478 |
-
# make copy of objects
|
479 |
-
objects = objects.copy()
|
480 |
-
|
481 |
-
# get the unique repoIds
|
482 |
-
objects["repoIdHash"] = objects.apply(_get_repo_id_with_hash, axis=1)
|
483 |
-
repo_id_hashes = set(objects["repoIdHash"].unique().tolist())
|
484 |
-
repo_ids = {
|
485 |
-
"/".join(repo_id_hash.split("/")[:2]) for repo_id_hash in repo_id_hashes
|
486 |
-
}
|
487 |
-
assert len(repo_id_hashes) == len(repo_ids), (
|
488 |
-
f"More than 1 commit hash per repoId!"
|
489 |
-
f" {len(repo_id_hashes)=}, {len(repo_ids)=}"
|
490 |
-
)
|
491 |
-
|
492 |
-
logger.info(
|
493 |
-
f"Provided {len(repo_ids)} repoIds with {len(objects)} objects to process."
|
494 |
-
)
|
495 |
-
|
496 |
-
# remove repoIds that have already been downloaded
|
497 |
-
repo_ids_to_download = repo_ids - downloaded_repo_ids
|
498 |
-
repo_id_hashes_to_download = [
|
499 |
-
repo_id_hash
|
500 |
-
for repo_id_hash in repo_id_hashes
|
501 |
-
if "/".join(repo_id_hash.split("/")[:2]) in repo_ids_to_download
|
502 |
-
]
|
503 |
-
|
504 |
-
logger.info(
|
505 |
-
f"Found {len(repo_ids_to_download)} repoIds not yet downloaded. Downloading now..."
|
506 |
-
)
|
507 |
-
|
508 |
-
# get the objects to download
|
509 |
-
groups = list(objects.groupby("repoIdHash"))
|
510 |
-
with Pool(processes=processes) as pool:
|
511 |
-
out_list = list(
|
512 |
-
tqdm(
|
513 |
-
pool.imap_unordered(_process_group, groups),
|
514 |
-
total=len(groups),
|
515 |
-
desc="Grouping objects by repository",
|
516 |
-
)
|
517 |
-
)
|
518 |
-
objects_per_repo_id_hash = dict(out_list)
|
519 |
-
|
520 |
-
all_args = [
|
521 |
(
|
522 |
repo_id_hash,
|
523 |
fs,
|
524 |
-
|
525 |
save_repo_format,
|
526 |
-
|
527 |
handle_found_object,
|
528 |
-
handle_missing_object,
|
529 |
handle_modified_object,
|
|
|
530 |
handle_new_object,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
531 |
)
|
532 |
-
|
533 |
-
|
534 |
-
|
535 |
-
|
536 |
-
|
537 |
-
|
538 |
-
|
539 |
-
|
540 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
541 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
542 |
)
|
543 |
-
|
544 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
10 |
from multiprocessing import Pool
|
11 |
from typing import Callable, Dict, List, Literal, Optional
|
12 |
from objaverse_xl.utils import get_file_hash
|
13 |
+
from objaverse_xl.abstract import ObjaverseSource
|
14 |
|
15 |
import fsspec
|
16 |
import pandas as pd
|
|
|
34 |
]
|
35 |
|
36 |
|
37 |
+
class GitHub(ObjaverseSource):
|
38 |
+
"""Script to download objects from GitHub."""
|
39 |
|
40 |
+
def load_annotations(self, download_dir: str = "~/.objaverse") -> pd.DataFrame:
|
41 |
+
"""Loads the GitHub 3D object metadata as a Pandas DataFrame.
|
|
|
42 |
|
43 |
+
Args:
|
44 |
+
download_dir (str, optional): Directory to download the parquet metadata
|
45 |
+
file. Supports all file systems supported by fsspec. Defaults to
|
46 |
+
"~/.objaverse".
|
|
|
|
|
|
|
47 |
|
48 |
+
Returns:
|
49 |
+
pd.DataFrame: GitHub 3D object metadata as a Pandas DataFrame with columns
|
50 |
+
for the object "fileIdentifier", "license", "source", "fileType",
|
51 |
+
"sha256", and "metadata".
|
52 |
+
"""
|
53 |
+
filename = os.path.join(download_dir, "github", "github-urls.parquet")
|
54 |
+
fs, path = fsspec.core.url_to_fs(filename)
|
55 |
+
fs.makedirs(os.path.dirname(path), exist_ok=True)
|
56 |
|
57 |
+
# download the parquet file if it doesn't exist
|
58 |
+
if not fs.exists(path):
|
59 |
+
url = "https://huggingface.co/datasets/allenai/objaverse-xl/resolve/main/github/github-urls.parquet"
|
|
|
60 |
|
61 |
+
response = requests.get(url)
|
62 |
+
response.raise_for_status()
|
63 |
+
with fs.open(path, "wb") as file:
|
64 |
+
file.write(response.content)
|
65 |
|
66 |
+
# load the parquet file with fsspec
|
67 |
+
with fs.open(path) as f:
|
68 |
+
df = pd.read_parquet(f)
|
69 |
|
70 |
+
df["metadata"] = "{}"
|
71 |
|
72 |
+
return df
|
73 |
|
74 |
+
def _get_repo_id_with_hash(self, item: pd.Series) -> str:
|
75 |
+
org, repo = item["fileIdentifier"].split("/")[3:5]
|
76 |
+
commit_hash = item["fileIdentifier"].split("/")[6]
|
77 |
+
return f"{org}/{repo}/{commit_hash}"
|
78 |
|
79 |
+
def _git_shallow_clone(self, repo_url: str, target_directory: str) -> bool:
|
80 |
+
"""Helper function to shallow clone a repo with git.
|
81 |
|
82 |
+
Args:
|
83 |
+
repo_url (str): URL of the repo to clone.
|
84 |
+
target_directory (str): Directory to clone the repo to.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
85 |
|
86 |
+
Returns:
|
87 |
+
bool: True if the clone was successful, False otherwise.
|
88 |
+
"""
|
89 |
+
return self._run_command_with_check(
|
90 |
+
["git", "clone", "--depth", "1", repo_url, target_directory],
|
91 |
+
)
|
92 |
|
93 |
+
def _run_command_with_check(
|
94 |
+
self, command: List[str], cwd: Optional[str] = None
|
95 |
+
) -> bool:
|
96 |
+
"""Helper function to run a command and check if it was successful.
|
97 |
+
|
98 |
+
Args:
|
99 |
+
command (List[str]): Command to run.
|
100 |
+
cwd (Optional[str], optional): Current working directory to run the command
|
101 |
+
in. Defaults to None.
|
102 |
+
|
103 |
+
Returns:
|
104 |
+
bool: True if the command was successful, False otherwise.
|
105 |
+
"""
|
106 |
+
try:
|
107 |
+
subprocess.run(
|
108 |
+
command,
|
109 |
+
cwd=cwd,
|
110 |
+
check=True,
|
111 |
+
stdout=subprocess.DEVNULL,
|
112 |
+
stderr=subprocess.DEVNULL,
|
113 |
+
)
|
114 |
+
return True
|
115 |
+
except subprocess.CalledProcessError as e:
|
116 |
+
logger.error("Error:", e)
|
117 |
+
logger.error(e.stdout)
|
118 |
+
logger.error(e.stderr)
|
119 |
+
return False
|
120 |
+
|
121 |
+
def _process_repo(
|
122 |
+
self,
|
123 |
+
repo_id: str,
|
124 |
+
fs: fsspec.AbstractFileSystem,
|
125 |
+
base_dir: str,
|
126 |
+
save_repo_format: Optional[Literal["zip", "tar", "tar.gz", "files"]],
|
127 |
+
expected_objects: Dict[str, str],
|
128 |
+
handle_found_object: Optional[Callable],
|
129 |
+
handle_modified_object: Optional[Callable],
|
130 |
+
handle_missing_object: Optional[Callable],
|
131 |
+
handle_new_object: Optional[Callable],
|
132 |
+
commit_hash: Optional[str],
|
133 |
+
) -> List[Dict[str, str]]:
|
134 |
+
"""Process a single repo.
|
135 |
+
|
136 |
+
Args:
|
137 |
+
repo_id (str): GitHub repo ID in the format of organization/repo.
|
138 |
+
fs (fsspec.AbstractFileSystem): File system to use for saving the repo.
|
139 |
+
base_dir (str): Base directory to save the repo to.
|
140 |
+
expected_objects (Dict[str, str]): Dictionary of objects that one expects to
|
141 |
+
find in the repo. Keys are the "fileIdentifier" (i.e., the GitHub URL in
|
142 |
+
this case) and values are the "sha256" of the objects.
|
143 |
+
{and the rest of the args are the same as download_github_objects}
|
144 |
+
|
145 |
+
Returns:
|
146 |
+
List[Dict[str, str]]: List of dictionaries with the keys "fileIdentifier"
|
147 |
+
and "sha256" for each downloaded object.
|
148 |
+
"""
|
149 |
+
# NOTE: assuming that the user has already checked that the repo doesn't exist,
|
150 |
+
org, repo = repo_id.split("/")
|
151 |
+
|
152 |
+
with tempfile.TemporaryDirectory() as temp_dir:
|
153 |
+
# clone the repo to a temp directory
|
154 |
+
target_directory = os.path.join(temp_dir, repo)
|
155 |
+
successful_clone = self._git_shallow_clone(
|
156 |
+
f"https://github.com/{org}/{repo}.git", target_directory
|
157 |
+
)
|
158 |
+
if not successful_clone:
|
159 |
+
logger.error(f"Could not clone {repo_id}")
|
160 |
+
if handle_missing_object is not None:
|
161 |
+
for github_url, sha256 in expected_objects.items():
|
162 |
+
handle_missing_object(
|
163 |
+
file_identifier=github_url,
|
164 |
+
sha256=sha256,
|
165 |
+
metadata=dict(github_organization=org, github_repo=repo),
|
166 |
+
)
|
167 |
+
return []
|
168 |
|
169 |
+
# use the commit hash if specified
|
170 |
+
repo_commit_hash = self._get_commit_hash_from_local_git_dir(
|
171 |
+
target_directory
|
172 |
+
)
|
173 |
+
if commit_hash is not None:
|
174 |
+
keep_going = True
|
175 |
+
if repo_commit_hash != commit_hash:
|
176 |
+
# run git reset --hard && git checkout 37f4d8d287e201ce52c048bf74d46d6a09d26b2c
|
177 |
+
if not self._run_command_with_check(
|
178 |
+
["git", "fetch", "origin", commit_hash], target_directory
|
179 |
+
):
|
180 |
+
logger.error(
|
181 |
+
f"Error in git fetch! Sticking with {repo_commit_hash=} instead of {commit_hash=}"
|
182 |
+
)
|
183 |
+
keep_going = False
|
184 |
|
185 |
+
if keep_going and not self._run_command_with_check(
|
186 |
+
["git", "reset", "--hard"], target_directory
|
187 |
+
):
|
188 |
+
logger.error(
|
189 |
+
f"Error in git reset! Sticking with {repo_commit_hash=} instead of {commit_hash=}"
|
190 |
+
)
|
191 |
+
keep_going = False
|
192 |
+
|
193 |
+
if keep_going:
|
194 |
+
if self._run_command_with_check(
|
195 |
+
["git", "checkout", commit_hash], target_directory
|
196 |
+
):
|
197 |
+
repo_commit_hash = commit_hash
|
198 |
+
else:
|
199 |
+
logger.error(
|
200 |
+
f"Error in git checkout! Sticking with {repo_commit_hash=} instead of {commit_hash=}"
|
201 |
+
)
|
202 |
+
|
203 |
+
# pull the lfs files
|
204 |
+
self._pull_lfs_files(target_directory)
|
205 |
+
|
206 |
+
# get all the files in the repo
|
207 |
+
files = self._list_files(target_directory)
|
208 |
+
files_with_3d_extension = [
|
209 |
+
file
|
210 |
+
for file in files
|
211 |
+
if any(file.lower().endswith(ext) for ext in FILE_EXTENSIONS)
|
212 |
+
]
|
213 |
+
|
214 |
+
# get the sha256 for each file
|
215 |
+
file_hashes = []
|
216 |
+
for file in tqdm(files_with_3d_extension, desc="Handling 3D object files"):
|
217 |
+
file_hash = get_file_hash(file)
|
218 |
+
# remove the temp_dir from the file path
|
219 |
+
github_url = file.replace(
|
220 |
+
target_directory,
|
221 |
+
f"https://github.com/{org}/{repo}/blob/{repo_commit_hash}",
|
222 |
+
)
|
223 |
+
file_hashes.append(dict(sha256=file_hash, fileIdentifier=github_url))
|
224 |
+
|
225 |
+
# handle the object under different conditions
|
226 |
+
if github_url in expected_objects:
|
227 |
+
if expected_objects[github_url] == file_hash:
|
228 |
+
if handle_found_object is not None:
|
229 |
+
handle_found_object(
|
230 |
+
local_path=file,
|
231 |
+
file_identifier=github_url,
|
232 |
+
sha256=file_hash,
|
233 |
+
metadata=dict(
|
234 |
+
github_organization=org, github_repo=repo
|
235 |
+
),
|
236 |
+
)
|
237 |
+
else:
|
238 |
+
if handle_modified_object is not None:
|
239 |
+
handle_modified_object(
|
240 |
+
local_path=file,
|
241 |
+
file_identifier=github_url,
|
242 |
+
new_sha256=file_hash,
|
243 |
+
old_sha256=expected_objects[github_url],
|
244 |
+
metadata=dict(
|
245 |
+
github_organization=org, github_repo=repo
|
246 |
+
),
|
247 |
+
)
|
248 |
+
elif handle_new_object is not None:
|
249 |
+
handle_new_object(
|
250 |
+
local_path=file,
|
251 |
+
file_identifier=github_url,
|
252 |
+
sha256=file_hash,
|
253 |
+
metadata=dict(github_organization=org, github_repo=repo),
|
254 |
+
)
|
255 |
|
256 |
+
# save the file hashes to a json file
|
257 |
+
with open(
|
258 |
+
os.path.join(target_directory, ".objaverse-file-hashes.json"),
|
259 |
+
"w",
|
260 |
+
encoding="utf-8",
|
261 |
+
) as f:
|
262 |
+
json.dump(file_hashes, f, indent=2)
|
263 |
+
|
264 |
+
# remove the .git directory
|
265 |
+
shutil.rmtree(os.path.join(target_directory, ".git"))
|
266 |
+
|
267 |
+
if save_repo_format is not None:
|
268 |
+
logger.debug(f"Saving as {save_repo_format}")
|
269 |
+
# save the repo to a zip file
|
270 |
+
if save_repo_format == "zip":
|
271 |
+
shutil.make_archive(target_directory, "zip", target_directory)
|
272 |
+
elif save_repo_format == "tar":
|
273 |
+
with tarfile.open(
|
274 |
+
os.path.join(temp_dir, f"{repo}.tar"), "w"
|
275 |
+
) as tar:
|
276 |
+
tar.add(target_directory, arcname=repo)
|
277 |
+
elif save_repo_format == "tar.gz":
|
278 |
+
with tarfile.open(
|
279 |
+
os.path.join(temp_dir, f"{repo}.tar.gz"), "w:gz"
|
280 |
+
) as tar:
|
281 |
+
tar.add(target_directory, arcname=repo)
|
282 |
+
elif save_repo_format == "files":
|
283 |
+
pass
|
284 |
+
else:
|
285 |
+
raise ValueError(
|
286 |
+
f"save_repo_format must be one of zip, tar, tar.gz, files. Got {save_repo_format}"
|
287 |
+
)
|
288 |
|
289 |
+
dirname = os.path.join(base_dir, "repos", org)
|
290 |
+
fs.makedirs(dirname, exist_ok=True)
|
291 |
+
if save_repo_format != "files":
|
292 |
+
# move the repo to the correct location (with put)
|
293 |
+
fs.put(
|
294 |
+
os.path.join(temp_dir, f"{repo}.{save_repo_format}"),
|
295 |
+
os.path.join(dirname, f"{repo}.{save_repo_format}"),
|
296 |
+
)
|
297 |
+
else:
|
298 |
+
# move the repo to the correct location (with put)
|
299 |
+
fs.put(target_directory, dirname, recursive=True)
|
300 |
+
|
301 |
+
# get each object that was missing from the expected objects
|
302 |
+
if handle_missing_object is not None:
|
303 |
+
obtained_urls = {x["fileIdentifier"] for x in file_hashes}
|
304 |
+
for github_url, sha256 in expected_objects.items():
|
305 |
+
if github_url not in obtained_urls:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
306 |
handle_missing_object(
|
307 |
file_identifier=github_url,
|
308 |
sha256=sha256,
|
309 |
metadata=dict(github_organization=org, github_repo=repo),
|
310 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
311 |
|
312 |
+
return file_hashes
|
|
|
|
|
|
|
|
|
|
|
|
|
313 |
|
314 |
+
def _list_files(self, root_dir: str) -> List[str]:
|
315 |
+
return [
|
316 |
+
os.path.join(root, f)
|
317 |
+
for root, dirs, files in os.walk(root_dir)
|
318 |
+
for f in files
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
319 |
]
|
320 |
|
321 |
+
def _pull_lfs_files(self, repo_dir: str) -> None:
|
322 |
+
if self._has_lfs_files(repo_dir):
|
323 |
+
subprocess.run(["git", "lfs", "pull"], cwd=repo_dir, check=True)
|
324 |
+
|
325 |
+
def _has_lfs_files(self, repo_dir: str) -> bool:
|
326 |
+
gitattributes_path = os.path.join(repo_dir, ".gitattributes")
|
327 |
+
if not os.path.exists(gitattributes_path):
|
328 |
+
return False
|
329 |
+
with open(gitattributes_path, "r", encoding="utf-8") as f:
|
330 |
+
for line in f:
|
331 |
+
if "filter=lfs" in line:
|
332 |
+
return True
|
333 |
+
return False
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
334 |
|
335 |
+
def _get_commit_hash_from_local_git_dir(self, local_git_dir: str) -> str:
|
336 |
+
# get the git hash of the repo
|
337 |
+
result = subprocess.run(
|
338 |
+
["git", "rev-parse", "HEAD"],
|
339 |
+
cwd=local_git_dir,
|
340 |
+
capture_output=True,
|
341 |
+
check=True,
|
342 |
+
)
|
343 |
+
commit_hash = result.stdout.strip().decode("utf-8")
|
344 |
+
return commit_hash
|
345 |
|
346 |
+
def _parallel_process_repo(self, args) -> List[Dict[str, str]]:
|
347 |
+
"""Helper function to process a repo in parallel.
|
|
|
|
|
348 |
|
349 |
+
Note: This function is used to parallelize the processing of repos. It is not
|
350 |
+
intended to be called directly.
|
351 |
|
352 |
+
Args:
|
353 |
+
args (Tuple): Tuple of arguments to pass to _process_repo.
|
|
|
354 |
|
355 |
+
Returns:
|
356 |
+
List[Dict[str, str]]: List of dictionaries with the keys "fileIdentifier"
|
357 |
+
and "sha256" for each downloaded object.
|
358 |
+
"""
|
359 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
360 |
(
|
361 |
repo_id_hash,
|
362 |
fs,
|
363 |
+
base_dir,
|
364 |
save_repo_format,
|
365 |
+
expected_objects,
|
366 |
handle_found_object,
|
|
|
367 |
handle_modified_object,
|
368 |
+
handle_missing_object,
|
369 |
handle_new_object,
|
370 |
+
) = args
|
371 |
+
repo_id = "/".join(repo_id_hash.split("/")[:2])
|
372 |
+
commit_hash = repo_id_hash.split("/")[2]
|
373 |
+
return self._process_repo(
|
374 |
+
repo_id=repo_id,
|
375 |
+
fs=fs,
|
376 |
+
base_dir=base_dir,
|
377 |
+
save_repo_format=save_repo_format,
|
378 |
+
expected_objects=expected_objects,
|
379 |
+
handle_found_object=handle_found_object,
|
380 |
+
handle_modified_object=handle_modified_object,
|
381 |
+
handle_missing_object=handle_missing_object,
|
382 |
+
handle_new_object=handle_new_object,
|
383 |
+
commit_hash=commit_hash,
|
384 |
)
|
385 |
+
|
386 |
+
def _process_group(self, group):
|
387 |
+
key, group_df = group
|
388 |
+
return key, group_df.set_index("fileIdentifier")["sha256"].to_dict()
|
389 |
+
|
390 |
+
def download_objects(
|
391 |
+
self,
|
392 |
+
objects: pd.DataFrame,
|
393 |
+
download_dir: str = "~/.objaverse",
|
394 |
+
processes: Optional[int] = None,
|
395 |
+
handle_found_object: Optional[Callable] = None,
|
396 |
+
handle_modified_object: Optional[Callable] = None,
|
397 |
+
handle_missing_object: Optional[Callable] = None,
|
398 |
+
*,
|
399 |
+
save_repo_format: Optional[Literal["zip", "tar", "tar.gz", "files"]] = None,
|
400 |
+
handle_new_object: Optional[Callable] = None,
|
401 |
+
**kwargs,
|
402 |
+
) -> List[Dict[str, str]]:
|
403 |
+
"""Download the specified GitHub objects.
|
404 |
+
|
405 |
+
Args:
|
406 |
+
objects (pd.DataFrmae): GitHub objects to download. Must have columns for
|
407 |
+
the object "fileIdentifier" and "sha256". Use the load_github_metadata
|
408 |
+
function to get the metadata.
|
409 |
+
download_dir (str, optional): Directory to download the GitHub objects to.
|
410 |
+
Supports all file systems supported by fsspec. Defaults to
|
411 |
+
"~/.objaverse".
|
412 |
+
processes (Optional[int], optional): Number of processes to use for
|
413 |
+
downloading. If None, will use the number of CPUs on the machine.
|
414 |
+
Defaults to None.
|
415 |
+
handle_found_object (Optional[Callable], optional): Called when an object is
|
416 |
+
successfully found and downloaded. Here, the object has the same sha256
|
417 |
+
as the one that was downloaded with Objaverse-XL. If None, the object
|
418 |
+
will be downloaded, but nothing will be done with it. Args for the
|
419 |
+
function include:
|
420 |
+
- local_path (str): Local path to the downloaded 3D object.
|
421 |
+
- file_identifier (str): GitHub URL of the 3D object.
|
422 |
+
- sha256 (str): SHA256 of the contents of the 3D object.
|
423 |
+
- metadata (Dict[str, Any]): Metadata about the 3D object, including the
|
424 |
+
GitHub organization and repo names.
|
425 |
+
Return is not used. Defaults to None.
|
426 |
+
handle_modified_object (Optional[Callable], optional): Called when a
|
427 |
+
modified object is found and downloaded. Here, the object is
|
428 |
+
successfully downloaded, but it has a different sha256 than the one that
|
429 |
+
was downloaded with Objaverse-XL. This is not expected to happen very
|
430 |
+
often, because the same commit hash is used for each repo. If None, the
|
431 |
+
object will be downloaded, but nothing will be done with it. Args for
|
432 |
+
the function include:
|
433 |
+
- local_path (str): Local path to the downloaded 3D object.
|
434 |
+
- file_identifier (str): GitHub URL of the 3D object.
|
435 |
+
- new_sha256 (str): SHA256 of the contents of the newly downloaded 3D
|
436 |
+
object.
|
437 |
+
- old_sha256 (str): Expected SHA256 of the contents of the 3D object as
|
438 |
+
it was when it was downloaded with Objaverse-XL.
|
439 |
+
- metadata (Dict[str, Any]): Metadata about the 3D object, including the
|
440 |
+
GitHub organization and repo names.
|
441 |
+
Return is not used. Defaults to None.
|
442 |
+
handle_missing_object (Optional[Callable], optional): Called when an object
|
443 |
+
that is in Objaverse-XL is not found. Here, it is likely that the
|
444 |
+
repository was deleted or renamed. If None, nothing will be done with
|
445 |
+
the missing object. Args for the function include:
|
446 |
+
- file_identifier (str): GitHub URL of the 3D object.
|
447 |
+
- sha256 (str): SHA256 of the contents of the original 3D object.
|
448 |
+
- metadata (Dict[str, Any]): Metadata about the 3D object, including the
|
449 |
+
GitHub organization and repo names.
|
450 |
+
Return is not used. Defaults to None.
|
451 |
+
save_repo_format (Optional[Literal["zip", "tar", "tar.gz", "files"]],
|
452 |
+
optional): Format to save the repository. If None, the repository will
|
453 |
+
not be saved. If "files" is specified, each file will be saved
|
454 |
+
individually. Otherwise, the repository can be saved as a "zip", "tar",
|
455 |
+
or "tar.gz" file. Defaults to None.
|
456 |
+
handle_new_object (Optional[Callable], optional): Called when a new object
|
457 |
+
is found. Here, the object is not used in Objaverse-XL, but is still
|
458 |
+
downloaded with the repository. The object may have not been used
|
459 |
+
because it does not successfully import into Blender. If None, the
|
460 |
+
object will be downloaded, but nothing will be done with it. Args for
|
461 |
+
the function include:
|
462 |
+
- local_path (str): Local path to the downloaded 3D object.
|
463 |
+
- file_identifier (str): GitHub URL of the 3D object.
|
464 |
+
- sha256 (str): SHA256 of the contents of the 3D object.
|
465 |
+
- metadata (Dict[str, Any]): Metadata about the 3D object, including the
|
466 |
+
GitHub organization and repo names.
|
467 |
+
Return is not used. Defaults to None.
|
468 |
+
|
469 |
+
Returns:
|
470 |
+
List[Dict[str, str]]: List of dictionaries with the keys "fileIdentifier"
|
471 |
+
and "sha256" for each downloaded object.
|
472 |
+
"""
|
473 |
+
if processes is None:
|
474 |
+
processes = multiprocessing.cpu_count()
|
475 |
+
|
476 |
+
base_download_dir = os.path.join(download_dir, "github")
|
477 |
+
fs, path = fsspec.core.url_to_fs(base_download_dir)
|
478 |
+
fs.makedirs(path, exist_ok=True)
|
479 |
+
|
480 |
+
# Getting immediate subdirectories of root_path
|
481 |
+
if save_repo_format == "files":
|
482 |
+
downloaded_repo_dirs = fs.glob(base_download_dir + "/repos/*/*/")
|
483 |
+
downloaded_repo_ids = {
|
484 |
+
"/".join(x.split("/")[-2:]) for x in downloaded_repo_dirs
|
485 |
+
}
|
486 |
+
else:
|
487 |
+
downloaded_repo_dirs = fs.glob(
|
488 |
+
base_download_dir + f"/repos/*/*.{save_repo_format}"
|
489 |
)
|
490 |
+
downloaded_repo_ids = set()
|
491 |
+
for x in downloaded_repo_dirs:
|
492 |
+
org, repo = x.split("/")[-2:]
|
493 |
+
repo = repo[: -len(f".{save_repo_format}")]
|
494 |
+
repo_id = f"{org}/{repo}"
|
495 |
+
downloaded_repo_ids.add(repo_id)
|
496 |
+
|
497 |
+
# make copy of objects
|
498 |
+
objects = objects.copy()
|
499 |
+
|
500 |
+
# get the unique repoIds
|
501 |
+
objects["repoIdHash"] = objects.apply(self._get_repo_id_with_hash, axis=1)
|
502 |
+
repo_id_hashes = set(objects["repoIdHash"].unique().tolist())
|
503 |
+
repo_ids = {
|
504 |
+
"/".join(repo_id_hash.split("/")[:2]) for repo_id_hash in repo_id_hashes
|
505 |
+
}
|
506 |
+
assert len(repo_id_hashes) == len(repo_ids), (
|
507 |
+
f"More than 1 commit hash per repoId!"
|
508 |
+
f" {len(repo_id_hashes)=}, {len(repo_ids)=}"
|
509 |
)
|
510 |
+
|
511 |
+
logger.info(
|
512 |
+
f"Provided {len(repo_ids)} repoIds with {len(objects)} objects to process."
|
513 |
+
)
|
514 |
+
|
515 |
+
# remove repoIds that have already been downloaded
|
516 |
+
repo_ids_to_download = repo_ids - downloaded_repo_ids
|
517 |
+
repo_id_hashes_to_download = [
|
518 |
+
repo_id_hash
|
519 |
+
for repo_id_hash in repo_id_hashes
|
520 |
+
if "/".join(repo_id_hash.split("/")[:2]) in repo_ids_to_download
|
521 |
+
]
|
522 |
+
|
523 |
+
logger.info(
|
524 |
+
f"Found {len(repo_ids_to_download)} repoIds not yet downloaded. Downloading now..."
|
525 |
+
)
|
526 |
+
|
527 |
+
# get the objects to download
|
528 |
+
groups = list(objects.groupby("repoIdHash"))
|
529 |
+
with Pool(processes=processes) as pool:
|
530 |
+
out_list = list(
|
531 |
+
tqdm(
|
532 |
+
pool.imap_unordered(self._process_group, groups),
|
533 |
+
total=len(groups),
|
534 |
+
desc="Grouping objects by repository",
|
535 |
+
)
|
536 |
+
)
|
537 |
+
objects_per_repo_id_hash = dict(out_list)
|
538 |
+
|
539 |
+
all_args = [
|
540 |
+
(
|
541 |
+
repo_id_hash,
|
542 |
+
fs,
|
543 |
+
path,
|
544 |
+
save_repo_format,
|
545 |
+
objects_per_repo_id_hash[repo_id_hash],
|
546 |
+
handle_found_object,
|
547 |
+
handle_missing_object,
|
548 |
+
handle_modified_object,
|
549 |
+
handle_new_object,
|
550 |
+
)
|
551 |
+
for repo_id_hash in repo_id_hashes_to_download
|
552 |
+
]
|
553 |
+
|
554 |
+
with Pool(processes=processes) as pool:
|
555 |
+
# use tqdm to show progress
|
556 |
+
out = list(
|
557 |
+
tqdm(
|
558 |
+
pool.imap_unordered(self._parallel_process_repo, all_args),
|
559 |
+
total=len(all_args),
|
560 |
+
)
|
561 |
+
)
|
562 |
+
out_list = [item for sublist in out for item in sublist]
|
563 |
+
return out_list
|