Datasets:

Languages:
English
ArXiv:
License:
mattdeitke's picture
update class names
ba96af0
raw
history blame
23.9 kB
"""Script to download objects from GitHub."""
import json
import multiprocessing
import os
import shutil
import subprocess
import tarfile
import tempfile
from multiprocessing import Pool
from typing import Callable, Dict, List, Literal, Optional
from objaverse_xl.utils import get_file_hash
from objaverse_xl.abstract import ObjaverseSource
import fsspec
import pandas as pd
import requests
from loguru import logger
from tqdm import tqdm
FILE_EXTENSIONS = [
".obj",
".glb",
".gltf",
".usdz",
".usd",
".fbx",
".stl",
".usda",
".dae",
".ply",
".abc",
".blend",
]
class GitHubDownloader(ObjaverseSource):
"""Script to download objects from GitHub."""
def load_annotations(self, download_dir: str = "~/.objaverse") -> pd.DataFrame:
"""Loads the GitHub 3D object metadata as a Pandas DataFrame.
Args:
download_dir (str, optional): Directory to download the parquet metadata
file. Supports all file systems supported by fsspec. Defaults to
"~/.objaverse".
Returns:
pd.DataFrame: GitHub 3D object metadata as a Pandas DataFrame with columns
for the object "fileIdentifier", "license", "source", "fileType",
"sha256", and "metadata".
"""
filename = os.path.join(download_dir, "github", "github-urls.parquet")
fs, path = fsspec.core.url_to_fs(filename)
fs.makedirs(os.path.dirname(path), exist_ok=True)
# download the parquet file if it doesn't exist
if not fs.exists(path):
url = "https://huggingface.co/datasets/allenai/objaverse-xl/resolve/main/github/github-urls.parquet"
response = requests.get(url)
response.raise_for_status()
with fs.open(path, "wb") as file:
file.write(response.content)
# load the parquet file with fsspec
with fs.open(path) as f:
df = pd.read_parquet(f)
df["metadata"] = "{}"
return df
def _get_repo_id_with_hash(self, item: pd.Series) -> str:
org, repo = item["fileIdentifier"].split("/")[3:5]
commit_hash = item["fileIdentifier"].split("/")[6]
return f"{org}/{repo}/{commit_hash}"
def _git_shallow_clone(self, repo_url: str, target_directory: str) -> bool:
"""Helper function to shallow clone a repo with git.
Args:
repo_url (str): URL of the repo to clone.
target_directory (str): Directory to clone the repo to.
Returns:
bool: True if the clone was successful, False otherwise.
"""
return self._run_command_with_check(
["git", "clone", "--depth", "1", repo_url, target_directory],
)
def _run_command_with_check(
self, command: List[str], cwd: Optional[str] = None
) -> bool:
"""Helper function to run a command and check if it was successful.
Args:
command (List[str]): Command to run.
cwd (Optional[str], optional): Current working directory to run the command
in. Defaults to None.
Returns:
bool: True if the command was successful, False otherwise.
"""
try:
subprocess.run(
command,
cwd=cwd,
check=True,
stdout=subprocess.DEVNULL,
stderr=subprocess.DEVNULL,
)
return True
except subprocess.CalledProcessError as e:
logger.error("Error:", e)
logger.error(e.stdout)
logger.error(e.stderr)
return False
def _process_repo(
self,
repo_id: str,
fs: fsspec.AbstractFileSystem,
base_dir: str,
save_repo_format: Optional[Literal["zip", "tar", "tar.gz", "files"]],
expected_objects: Dict[str, str],
handle_found_object: Optional[Callable],
handle_modified_object: Optional[Callable],
handle_missing_object: Optional[Callable],
handle_new_object: Optional[Callable],
commit_hash: Optional[str],
) -> List[Dict[str, str]]:
"""Process a single repo.
Args:
repo_id (str): GitHub repo ID in the format of organization/repo.
fs (fsspec.AbstractFileSystem): File system to use for saving the repo.
base_dir (str): Base directory to save the repo to.
expected_objects (Dict[str, str]): Dictionary of objects that one expects to
find in the repo. Keys are the "fileIdentifier" (i.e., the GitHub URL in
this case) and values are the "sha256" of the objects.
{and the rest of the args are the same as download_github_objects}
Returns:
List[Dict[str, str]]: List of dictionaries with the keys "fileIdentifier"
and "sha256" for each downloaded object.
"""
# NOTE: assuming that the user has already checked that the repo doesn't exist,
org, repo = repo_id.split("/")
with tempfile.TemporaryDirectory() as temp_dir:
# clone the repo to a temp directory
target_directory = os.path.join(temp_dir, repo)
successful_clone = self._git_shallow_clone(
f"https://github.com/{org}/{repo}.git", target_directory
)
if not successful_clone:
logger.error(f"Could not clone {repo_id}")
if handle_missing_object is not None:
for github_url, sha256 in expected_objects.items():
handle_missing_object(
file_identifier=github_url,
sha256=sha256,
metadata=dict(github_organization=org, github_repo=repo),
)
return []
# use the commit hash if specified
repo_commit_hash = self._get_commit_hash_from_local_git_dir(
target_directory
)
if commit_hash is not None:
keep_going = True
if repo_commit_hash != commit_hash:
# run git reset --hard && git checkout 37f4d8d287e201ce52c048bf74d46d6a09d26b2c
if not self._run_command_with_check(
["git", "fetch", "origin", commit_hash], target_directory
):
logger.error(
f"Error in git fetch! Sticking with {repo_commit_hash=} instead of {commit_hash=}"
)
keep_going = False
if keep_going and not self._run_command_with_check(
["git", "reset", "--hard"], target_directory
):
logger.error(
f"Error in git reset! Sticking with {repo_commit_hash=} instead of {commit_hash=}"
)
keep_going = False
if keep_going:
if self._run_command_with_check(
["git", "checkout", commit_hash], target_directory
):
repo_commit_hash = commit_hash
else:
logger.error(
f"Error in git checkout! Sticking with {repo_commit_hash=} instead of {commit_hash=}"
)
# pull the lfs files
self._pull_lfs_files(target_directory)
# get all the files in the repo
files = self._list_files(target_directory)
files_with_3d_extension = [
file
for file in files
if any(file.lower().endswith(ext) for ext in FILE_EXTENSIONS)
]
# get the sha256 for each file
file_hashes = []
for file in tqdm(files_with_3d_extension, desc="Handling 3D object files"):
file_hash = get_file_hash(file)
# remove the temp_dir from the file path
github_url = file.replace(
target_directory,
f"https://github.com/{org}/{repo}/blob/{repo_commit_hash}",
)
file_hashes.append(dict(sha256=file_hash, fileIdentifier=github_url))
# handle the object under different conditions
if github_url in expected_objects:
if expected_objects[github_url] == file_hash:
if handle_found_object is not None:
handle_found_object(
local_path=file,
file_identifier=github_url,
sha256=file_hash,
metadata=dict(
github_organization=org, github_repo=repo
),
)
else:
if handle_modified_object is not None:
handle_modified_object(
local_path=file,
file_identifier=github_url,
new_sha256=file_hash,
old_sha256=expected_objects[github_url],
metadata=dict(
github_organization=org, github_repo=repo
),
)
elif handle_new_object is not None:
handle_new_object(
local_path=file,
file_identifier=github_url,
sha256=file_hash,
metadata=dict(github_organization=org, github_repo=repo),
)
# save the file hashes to a json file
with open(
os.path.join(target_directory, ".objaverse-file-hashes.json"),
"w",
encoding="utf-8",
) as f:
json.dump(file_hashes, f, indent=2)
# remove the .git directory
shutil.rmtree(os.path.join(target_directory, ".git"))
if save_repo_format is not None:
logger.debug(f"Saving as {save_repo_format}")
# save the repo to a zip file
if save_repo_format == "zip":
shutil.make_archive(target_directory, "zip", target_directory)
elif save_repo_format == "tar":
with tarfile.open(
os.path.join(temp_dir, f"{repo}.tar"), "w"
) as tar:
tar.add(target_directory, arcname=repo)
elif save_repo_format == "tar.gz":
with tarfile.open(
os.path.join(temp_dir, f"{repo}.tar.gz"), "w:gz"
) as tar:
tar.add(target_directory, arcname=repo)
elif save_repo_format == "files":
pass
else:
raise ValueError(
f"save_repo_format must be one of zip, tar, tar.gz, files. Got {save_repo_format}"
)
dirname = os.path.join(base_dir, "repos", org)
fs.makedirs(dirname, exist_ok=True)
if save_repo_format != "files":
# move the repo to the correct location (with put)
fs.put(
os.path.join(temp_dir, f"{repo}.{save_repo_format}"),
os.path.join(dirname, f"{repo}.{save_repo_format}"),
)
else:
# move the repo to the correct location (with put)
fs.put(target_directory, dirname, recursive=True)
# get each object that was missing from the expected objects
if handle_missing_object is not None:
obtained_urls = {x["fileIdentifier"] for x in file_hashes}
for github_url, sha256 in expected_objects.items():
if github_url not in obtained_urls:
handle_missing_object(
file_identifier=github_url,
sha256=sha256,
metadata=dict(github_organization=org, github_repo=repo),
)
return file_hashes
def _list_files(self, root_dir: str) -> List[str]:
return [
os.path.join(root, f)
for root, dirs, files in os.walk(root_dir)
for f in files
]
def _pull_lfs_files(self, repo_dir: str) -> None:
if self._has_lfs_files(repo_dir):
subprocess.run(["git", "lfs", "pull"], cwd=repo_dir, check=True)
def _has_lfs_files(self, repo_dir: str) -> bool:
gitattributes_path = os.path.join(repo_dir, ".gitattributes")
if not os.path.exists(gitattributes_path):
return False
with open(gitattributes_path, "r", encoding="utf-8") as f:
for line in f:
if "filter=lfs" in line:
return True
return False
def _get_commit_hash_from_local_git_dir(self, local_git_dir: str) -> str:
# get the git hash of the repo
result = subprocess.run(
["git", "rev-parse", "HEAD"],
cwd=local_git_dir,
capture_output=True,
check=True,
)
commit_hash = result.stdout.strip().decode("utf-8")
return commit_hash
def _parallel_process_repo(self, args) -> List[Dict[str, str]]:
"""Helper function to process a repo in parallel.
Note: This function is used to parallelize the processing of repos. It is not
intended to be called directly.
Args:
args (Tuple): Tuple of arguments to pass to _process_repo.
Returns:
List[Dict[str, str]]: List of dictionaries with the keys "fileIdentifier"
and "sha256" for each downloaded object.
"""
(
repo_id_hash,
fs,
base_dir,
save_repo_format,
expected_objects,
handle_found_object,
handle_modified_object,
handle_missing_object,
handle_new_object,
) = args
repo_id = "/".join(repo_id_hash.split("/")[:2])
commit_hash = repo_id_hash.split("/")[2]
return self._process_repo(
repo_id=repo_id,
fs=fs,
base_dir=base_dir,
save_repo_format=save_repo_format,
expected_objects=expected_objects,
handle_found_object=handle_found_object,
handle_modified_object=handle_modified_object,
handle_missing_object=handle_missing_object,
handle_new_object=handle_new_object,
commit_hash=commit_hash,
)
def _process_group(self, group):
key, group_df = group
return key, group_df.set_index("fileIdentifier")["sha256"].to_dict()
def download_objects(
self,
objects: pd.DataFrame,
download_dir: str = "~/.objaverse",
processes: Optional[int] = None,
handle_found_object: Optional[Callable] = None,
handle_modified_object: Optional[Callable] = None,
handle_missing_object: Optional[Callable] = None,
*,
save_repo_format: Optional[Literal["zip", "tar", "tar.gz", "files"]] = None,
handle_new_object: Optional[Callable] = None,
**kwargs,
) -> List[Dict[str, str]]:
"""Download the specified GitHub objects.
Args:
objects (pd.DataFrmae): GitHub objects to download. Must have columns for
the object "fileIdentifier" and "sha256". Use the load_github_metadata
function to get the metadata.
download_dir (str, optional): Directory to download the GitHub objects to.
Supports all file systems supported by fsspec. Defaults to
"~/.objaverse".
processes (Optional[int], optional): Number of processes to use for
downloading. If None, will use the number of CPUs on the machine.
Defaults to None.
handle_found_object (Optional[Callable], optional): Called when an object is
successfully found and downloaded. Here, the object has the same sha256
as the one that was downloaded with Objaverse-XL. If None, the object
will be downloaded, but nothing will be done with it. Args for the
function include:
- local_path (str): Local path to the downloaded 3D object.
- file_identifier (str): GitHub URL of the 3D object.
- sha256 (str): SHA256 of the contents of the 3D object.
- metadata (Dict[str, Any]): Metadata about the 3D object, including the
GitHub organization and repo names.
Return is not used. Defaults to None.
handle_modified_object (Optional[Callable], optional): Called when a
modified object is found and downloaded. Here, the object is
successfully downloaded, but it has a different sha256 than the one that
was downloaded with Objaverse-XL. This is not expected to happen very
often, because the same commit hash is used for each repo. If None, the
object will be downloaded, but nothing will be done with it. Args for
the function include:
- local_path (str): Local path to the downloaded 3D object.
- file_identifier (str): GitHub URL of the 3D object.
- new_sha256 (str): SHA256 of the contents of the newly downloaded 3D
object.
- old_sha256 (str): Expected SHA256 of the contents of the 3D object as
it was when it was downloaded with Objaverse-XL.
- metadata (Dict[str, Any]): Metadata about the 3D object, including the
GitHub organization and repo names.
Return is not used. Defaults to None.
handle_missing_object (Optional[Callable], optional): Called when an object
that is in Objaverse-XL is not found. Here, it is likely that the
repository was deleted or renamed. If None, nothing will be done with
the missing object. Args for the function include:
- file_identifier (str): GitHub URL of the 3D object.
- sha256 (str): SHA256 of the contents of the original 3D object.
- metadata (Dict[str, Any]): Metadata about the 3D object, including the
GitHub organization and repo names.
Return is not used. Defaults to None.
save_repo_format (Optional[Literal["zip", "tar", "tar.gz", "files"]],
optional): Format to save the repository. If None, the repository will
not be saved. If "files" is specified, each file will be saved
individually. Otherwise, the repository can be saved as a "zip", "tar",
or "tar.gz" file. Defaults to None.
handle_new_object (Optional[Callable], optional): Called when a new object
is found. Here, the object is not used in Objaverse-XL, but is still
downloaded with the repository. The object may have not been used
because it does not successfully import into Blender. If None, the
object will be downloaded, but nothing will be done with it. Args for
the function include:
- local_path (str): Local path to the downloaded 3D object.
- file_identifier (str): GitHub URL of the 3D object.
- sha256 (str): SHA256 of the contents of the 3D object.
- metadata (Dict[str, Any]): Metadata about the 3D object, including the
GitHub organization and repo names.
Return is not used. Defaults to None.
Returns:
List[Dict[str, str]]: List of dictionaries with the keys "fileIdentifier"
and "sha256" for each downloaded object.
"""
if processes is None:
processes = multiprocessing.cpu_count()
base_download_dir = os.path.join(download_dir, "github")
fs, path = fsspec.core.url_to_fs(base_download_dir)
fs.makedirs(path, exist_ok=True)
# Getting immediate subdirectories of root_path
if save_repo_format == "files":
downloaded_repo_dirs = fs.glob(base_download_dir + "/repos/*/*/")
downloaded_repo_ids = {
"/".join(x.split("/")[-2:]) for x in downloaded_repo_dirs
}
else:
downloaded_repo_dirs = fs.glob(
base_download_dir + f"/repos/*/*.{save_repo_format}"
)
downloaded_repo_ids = set()
for x in downloaded_repo_dirs:
org, repo = x.split("/")[-2:]
repo = repo[: -len(f".{save_repo_format}")]
repo_id = f"{org}/{repo}"
downloaded_repo_ids.add(repo_id)
# make copy of objects
objects = objects.copy()
# get the unique repoIds
objects["repoIdHash"] = objects.apply(self._get_repo_id_with_hash, axis=1)
repo_id_hashes = set(objects["repoIdHash"].unique().tolist())
repo_ids = {
"/".join(repo_id_hash.split("/")[:2]) for repo_id_hash in repo_id_hashes
}
assert len(repo_id_hashes) == len(repo_ids), (
f"More than 1 commit hash per repoId!"
f" {len(repo_id_hashes)=}, {len(repo_ids)=}"
)
logger.info(
f"Provided {len(repo_ids)} repoIds with {len(objects)} objects to process."
)
# remove repoIds that have already been downloaded
repo_ids_to_download = repo_ids - downloaded_repo_ids
repo_id_hashes_to_download = [
repo_id_hash
for repo_id_hash in repo_id_hashes
if "/".join(repo_id_hash.split("/")[:2]) in repo_ids_to_download
]
logger.info(
f"Found {len(repo_ids_to_download)} repoIds not yet downloaded. Downloading now..."
)
# get the objects to download
groups = list(objects.groupby("repoIdHash"))
with Pool(processes=processes) as pool:
out_list = list(
tqdm(
pool.imap_unordered(self._process_group, groups),
total=len(groups),
desc="Grouping objects by repository",
)
)
objects_per_repo_id_hash = dict(out_list)
all_args = [
(
repo_id_hash,
fs,
path,
save_repo_format,
objects_per_repo_id_hash[repo_id_hash],
handle_found_object,
handle_missing_object,
handle_modified_object,
handle_new_object,
)
for repo_id_hash in repo_id_hashes_to_download
]
with Pool(processes=processes) as pool:
# use tqdm to show progress
out = list(
tqdm(
pool.imap_unordered(self._parallel_process_repo, all_args),
total=len(all_args),
)
)
out_list = [item for sublist in out for item in sublist]
return out_list