Datasets:
File size: 5,664 Bytes
47652a2 | 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 | """Manifest-validated Hugging Face upload with SDK-managed concurrency and resume."""
import argparse
from concurrent.futures import ThreadPoolExecutor
import json
import logging
import os
from pathlib import Path
from filelock import FileLock
from huggingface_hub import HfApi
from huggingface_hub.errors import RepositoryNotFoundError
from convert_dataset import sha256, save_json
def main():
p = argparse.ArgumentParser(description=__doc__)
p.add_argument('--folder', type=Path, required=True)
p.add_argument('--repo-id', default='leeibo/TraceSpatial-Trace')
p.add_argument('--workers', type=int, default=4, help='Checksum workers; also upload workers when --legacy-workers is used')
p.add_argument('--execute', action='store_true', help='Without this flag, only validate and show upload plan')
p.add_argument('--legacy-workers', action='store_true', help='Use resumable upload_large_folder for explicit worker control (deprecated upstream)')
args = p.parse_args()
if args.workers < 1: p.error('--workers must be positive')
root = args.folder.resolve()
manifest_path = root/'manifest.json'
if not manifest_path.is_file(): p.error('Conversion manifest.json missing; finish conversion first')
manifest = json.loads(manifest_path.read_text(encoding='utf-8'))
if not manifest.get('complete'): p.error('Conversion not complete')
(root/'.state').mkdir(exist_ok=True)
logging.basicConfig(level=logging.INFO, format='%(asctime)s %(message)s',
handlers=[logging.StreamHandler(), logging.FileHandler(root/'.state'/'upload.log', encoding='utf-8')])
with FileLock(root/'.state'/'upload.lock', timeout=0), FileLock(root/'.state'/'convert.lock', timeout=0):
def validate(item):
path = (root/item['path']).resolve()
if root not in path.parents or not path.is_file():
raise RuntimeError(f'Invalid/missing shard path: {item["path"]}')
if path.stat().st_size != item['bytes'] or sha256(path) != item['sha256']:
raise RuntimeError(f'Shard checksum mismatch: {item["path"]}')
with ThreadPoolExecutor(max_workers=args.workers) as pool:
list(pool.map(validate, manifest['files']))
tracked = {x['path'] for x in manifest['files']}
actual = {f.relative_to(root).as_posix() for f in root.glob('data/**/*.parquet')}
if tracked != actual:
raise RuntimeError('Untracked or missing data shards; use a clean output directory')
if not (root/'README.md').is_file(): raise RuntimeError('README.md missing')
ancillary = [p.relative_to(root).as_posix() for pattern in ('scripts/*.py','scripts/*.md','scripts/*.txt','scripts/*.yml','assets/*') for p in root.glob(pattern) if p.is_file()]
allow = sorted(tracked) + ['README.md','manifest.json'] + ancillary
logging.info('Plan: %s files, %.3f GiB in data shards, %s QA rows -> %s',len(allow),sum(f['bytes'] for f in manifest['files'])/1024**3,manifest['qa_rows'],args.repo_id)
if not args.execute:
logging.info('Dry run complete; add --execute to upload. No network writes performed.')
return
# Bind SDK resume cache to one destination. Never print or embed the token.
binding = root/'.state'/'upload-target.json'
target = {'repo_id':args.repo_id, 'repo_type':'dataset'}
if binding.exists() and json.loads(binding.read_text(encoding='utf-8')) != target:
raise RuntimeError('This folder upload cache belongs to another repo; use a fresh staging folder')
api = HfApi()
logging.info('Authenticated as %s',api.whoami()['name'])
try:
info = api.dataset_info(args.repo_id)
logging.info('Existing repository visibility: %s', 'private' if info.private else 'public')
except RepositoryNotFoundError:
api.create_repo(args.repo_id, repo_type='dataset', private=True, exist_ok=True)
save_json(binding,target)
if args.legacy_workers:
# Explicitly requested compatibility mode; no concurrent committing processes.
api.upload_large_folder(repo_id=args.repo_id, repo_type='dataset', folder_path=root,
num_workers=args.workers, allow_patterns=allow, print_report_every=30)
else:
from huggingface_hub.utils._runtime import is_xet_available
if not is_xet_available():
raise RuntimeError("Install hf-xet from requirements.txt for resumable multi-commit upload")
# Current SDK: internally concurrent, resumable, automatically splits commits.
api.upload_folder(repo_id=args.repo_id, repo_type='dataset', folder_path=root,
allow_patterns=allow, commit_message='Publish TraceSpatial RGB and QA Parquet release')
remote = {f.path:f for f in api.list_repo_tree(args.repo_id,repo_type='dataset',recursive=True) if hasattr(f,'size')}
for item in manifest['files']:
f = remote.get(item['path'])
if f is None or f.size != item['bytes']:
raise RuntimeError(f'Remote size mismatch: {item["path"]}')
lfs = getattr(f,'lfs',None)
remote_sha = getattr(lfs,'sha256',None)
if remote_sha and remote_sha != item['sha256']:
raise RuntimeError(f'Remote SHA-256 mismatch: {item["path"]}')
logging.info('Upload verified: https://huggingface.co/datasets/%s',args.repo_id)
if __name__ == '__main__':
main()
|