Upload upload.py with huggingface_hub
Browse files
upload.py
ADDED
|
@@ -0,0 +1,297 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#!/usr/bin/env python
|
| 2 |
+
"""
|
| 3 |
+
upload.py
|
| 4 |
+
|
| 5 |
+
Upload the QMCSoftware/LDData repository (low discrepancy generating vectors
|
| 6 |
+
and matrices) to the Hugging Face Datasets Hub as a dataset repo.
|
| 7 |
+
|
| 8 |
+
This script:
|
| 9 |
+
|
| 10 |
+
1. Creates (or reuses) a dataset repo on the Hub.
|
| 11 |
+
2. Uploads all files from a local LDData checkout using `upload_folder`.
|
| 12 |
+
3. Leaves README.md in place so it becomes the dataset card.
|
| 13 |
+
After upload you can edit the card in the web UI to:
|
| 14 |
+
- Link to the paper (arXiv:2502.14256).
|
| 15 |
+
- Add "Citation" and "Uses" sections, similar to
|
| 16 |
+
- facebook/omnilingual-asr-corpus
|
| 17 |
+
- nvidia/PhysicalAI-Autonomous-Vehicles
|
| 18 |
+
- moondream/refcoco-m
|
| 19 |
+
|
| 20 |
+
Requirements:
|
| 21 |
+
pip install "huggingface_hub>=0.32.0"
|
| 22 |
+
|
| 23 |
+
Authentication:
|
| 24 |
+
- Either set HF_TOKEN in your environment:
|
| 25 |
+
export HF_TOKEN=hf_xxx...
|
| 26 |
+
OR pass --token on the command line.
|
| 27 |
+
|
| 28 |
+
Example usage:
|
| 29 |
+
python upload.py \
|
| 30 |
+
--repo-id QMCSoftware/LDData \
|
| 31 |
+
--local-path /path/to/local/LDData
|
| 32 |
+
|
| 33 |
+
After upload you’ll be able to do, e.g.:
|
| 34 |
+
|
| 35 |
+
from datasets import load_dataset
|
| 36 |
+
ds = load_dataset("QMCSoftware/LDData")
|
| 37 |
+
|
| 38 |
+
and link the dataset to your paper page on Hugging Face.
|
| 39 |
+
"""
|
| 40 |
+
|
| 41 |
+
import argparse
|
| 42 |
+
import os
|
| 43 |
+
import sys
|
| 44 |
+
from pathlib import Path
|
| 45 |
+
import httpx
|
| 46 |
+
import fnmatch
|
| 47 |
+
|
| 48 |
+
from huggingface_hub import HfApi, create_repo # type: ignore
|
| 49 |
+
|
| 50 |
+
|
| 51 |
+
def parse_args() -> argparse.Namespace:
|
| 52 |
+
parser = argparse.ArgumentParser(
|
| 53 |
+
description="Upload LDData to the Hugging Face Datasets Hub."
|
| 54 |
+
)
|
| 55 |
+
parser.add_argument(
|
| 56 |
+
"--repo-id",
|
| 57 |
+
type=str,
|
| 58 |
+
default="QMCSoftware/LDData",
|
| 59 |
+
help="Target dataset repo id on the Hub (e.g. 'QMCSoftware/LDData').",
|
| 60 |
+
)
|
| 61 |
+
parser.add_argument(
|
| 62 |
+
"--local-path",
|
| 63 |
+
type=str,
|
| 64 |
+
default=".",
|
| 65 |
+
help="Path to local LDData checkout (default: current directory).",
|
| 66 |
+
)
|
| 67 |
+
parser.add_argument(
|
| 68 |
+
"--token",
|
| 69 |
+
type=str,
|
| 70 |
+
default=None,
|
| 71 |
+
help="Hugging Face access token. If omitted, HF_TOKEN env var is used.",
|
| 72 |
+
)
|
| 73 |
+
parser.add_argument(
|
| 74 |
+
"--private",
|
| 75 |
+
action="store_true",
|
| 76 |
+
help="Create the dataset repo as private (default: public).",
|
| 77 |
+
)
|
| 78 |
+
parser.add_argument(
|
| 79 |
+
"--dry-run",
|
| 80 |
+
action="store_true",
|
| 81 |
+
help="Do not upload, just print what would be done.",
|
| 82 |
+
)
|
| 83 |
+
return parser.parse_args()
|
| 84 |
+
|
| 85 |
+
|
| 86 |
+
def get_token(cmd_token: str | None) -> str:
|
| 87 |
+
token = cmd_token or os.environ.get("HF_TOKEN")
|
| 88 |
+
if not token:
|
| 89 |
+
raise SystemExit(
|
| 90 |
+
"No token provided. Please either:\n"
|
| 91 |
+
" - set HF_TOKEN in your environment, or\n"
|
| 92 |
+
" - pass --token hf_xxx... on the command line."
|
| 93 |
+
)
|
| 94 |
+
return token
|
| 95 |
+
|
| 96 |
+
|
| 97 |
+
def main() -> None:
|
| 98 |
+
args = parse_args()
|
| 99 |
+
|
| 100 |
+
local_path = Path(args.local_path).expanduser().resolve()
|
| 101 |
+
if not local_path.exists():
|
| 102 |
+
raise SystemExit(f"Local path does not exist: {local_path}")
|
| 103 |
+
|
| 104 |
+
# Sanity check: are we in LDData?
|
| 105 |
+
readme = local_path / "README.md"
|
| 106 |
+
if not readme.exists():
|
| 107 |
+
print(
|
| 108 |
+
f"WARNING: {readme} does not exist. "
|
| 109 |
+
"Are you sure this is the LDData repo root?",
|
| 110 |
+
file=sys.stderr,
|
| 111 |
+
)
|
| 112 |
+
|
| 113 |
+
token = get_token(args.token)
|
| 114 |
+
repo_id = args.repo_id
|
| 115 |
+
|
| 116 |
+
print(f"Using repo_id: {repo_id}")
|
| 117 |
+
print(f"Local path : {local_path}")
|
| 118 |
+
print(f"Private : {args.private}")
|
| 119 |
+
if args.dry_run:
|
| 120 |
+
print("Dry run enabled: NOT creating or uploading, just showing intent.")
|
| 121 |
+
return
|
| 122 |
+
|
| 123 |
+
# Initialize API client
|
| 124 |
+
api = HfApi(token=token)
|
| 125 |
+
|
| 126 |
+
# 1. Create (or reuse) the dataset repo on the Hub
|
| 127 |
+
print(f"Creating (or reusing) dataset repo '{repo_id}' on the Hub...")
|
| 128 |
+
create_repo(
|
| 129 |
+
repo_id=repo_id,
|
| 130 |
+
repo_type="dataset",
|
| 131 |
+
private=args.private,
|
| 132 |
+
exist_ok=True,
|
| 133 |
+
token=token,
|
| 134 |
+
)
|
| 135 |
+
|
| 136 |
+
# 2. Upload folder contents
|
| 137 |
+
# We ignore some typical non-data files to keep the repo clean.
|
| 138 |
+
# Adjust this list if you want to exclude more or fewer things.
|
| 139 |
+
ignore_patterns = [
|
| 140 |
+
".git/*",
|
| 141 |
+
".gitignore",
|
| 142 |
+
".DS_Store",
|
| 143 |
+
"__pycache__/*",
|
| 144 |
+
"*.pyc",
|
| 145 |
+
"*.pyo",
|
| 146 |
+
"*~",
|
| 147 |
+
"*.ipynb_checkpoints*",
|
| 148 |
+
# If you do NOT want to upload the demo notebook or env file, keep these:
|
| 149 |
+
# "LDData Demo.ipynb",
|
| 150 |
+
# "env.yml",
|
| 151 |
+
]
|
| 152 |
+
|
| 153 |
+
print("Uploading local folder to the Hub (this may take a while)...")
|
| 154 |
+
|
| 155 |
+
# Build list of candidate files and compute total size (exclude ignored)
|
| 156 |
+
def is_ignored(rel_path: str) -> bool:
|
| 157 |
+
# fnmatch against each ignore pattern; use posix style paths
|
| 158 |
+
for patt in ignore_patterns:
|
| 159 |
+
if fnmatch.fnmatch(rel_path, patt):
|
| 160 |
+
return True
|
| 161 |
+
# also allow matching against patterns that start with a leading slash
|
| 162 |
+
if fnmatch.fnmatch("/" + rel_path, patt):
|
| 163 |
+
return True
|
| 164 |
+
return False
|
| 165 |
+
|
| 166 |
+
files_to_upload = []
|
| 167 |
+
total_size = 0
|
| 168 |
+
for root, _, files in os.walk(local_path):
|
| 169 |
+
for fname in files:
|
| 170 |
+
full = Path(root) / fname
|
| 171 |
+
try:
|
| 172 |
+
rel = full.relative_to(local_path).as_posix()
|
| 173 |
+
except Exception:
|
| 174 |
+
rel = full.as_posix()
|
| 175 |
+
if is_ignored(rel):
|
| 176 |
+
continue
|
| 177 |
+
try:
|
| 178 |
+
sz = full.stat().st_size
|
| 179 |
+
except OSError:
|
| 180 |
+
sz = 0
|
| 181 |
+
files_to_upload.append((full, rel))
|
| 182 |
+
total_size += sz
|
| 183 |
+
|
| 184 |
+
# Threshold to prefer upload_large_folder (50 MiB)
|
| 185 |
+
LARGE_THRESHOLD = 50 * 1024 * 1024
|
| 186 |
+
|
| 187 |
+
try:
|
| 188 |
+
if total_size > LARGE_THRESHOLD and hasattr(api, "upload_large_folder"):
|
| 189 |
+
# Prefer API method designed for large folders when available.
|
| 190 |
+
try:
|
| 191 |
+
# Try calling with ignore_patterns if supported
|
| 192 |
+
api.upload_large_folder(
|
| 193 |
+
folder_path=str(local_path),
|
| 194 |
+
repo_id=repo_id,
|
| 195 |
+
repo_type="dataset",
|
| 196 |
+
ignore_patterns=ignore_patterns,
|
| 197 |
+
token=token,
|
| 198 |
+
)
|
| 199 |
+
except TypeError:
|
| 200 |
+
# Fallback: call without ignore_patterns if signature differs
|
| 201 |
+
try:
|
| 202 |
+
api.upload_large_folder(
|
| 203 |
+
folder_path=str(local_path),
|
| 204 |
+
repo_id=repo_id,
|
| 205 |
+
repo_type="dataset",
|
| 206 |
+
token=token,
|
| 207 |
+
)
|
| 208 |
+
except Exception as exc:
|
| 209 |
+
print()
|
| 210 |
+
print("ERROR: upload_large_folder failed:")
|
| 211 |
+
print(f" {exc!r}")
|
| 212 |
+
print("Falling back to per-file upload...")
|
| 213 |
+
# fall through to per-file upload below
|
| 214 |
+
raise
|
| 215 |
+
else:
|
| 216 |
+
# For small-ish folders or when upload_large_folder is not available,
|
| 217 |
+
# use upload_folder. This may raise httpx.ReadTimeout which we handle.
|
| 218 |
+
api.upload_folder(
|
| 219 |
+
folder_path=str(local_path),
|
| 220 |
+
repo_id=repo_id,
|
| 221 |
+
repo_type="dataset",
|
| 222 |
+
ignore_patterns=ignore_patterns,
|
| 223 |
+
)
|
| 224 |
+
except httpx.ReadTimeout:
|
| 225 |
+
print()
|
| 226 |
+
print("ERROR: Upload read timed out.")
|
| 227 |
+
print("Possible actions:")
|
| 228 |
+
print(" - Check your network connection and try again.")
|
| 229 |
+
print(" - Try uploading in smaller batches (split large files or directories).")
|
| 230 |
+
print(" - Use `HfApi().upload_large_folder(...)` or the CLI `hf upload-large-folder` if available.")
|
| 231 |
+
print(" - Upgrade huggingface_hub to the latest version in case it adds improved timeout handling.")
|
| 232 |
+
print(" - If you have very large files, consider using git-lfs or the web UI.")
|
| 233 |
+
sys.exit(1)
|
| 234 |
+
except TypeError as exc:
|
| 235 |
+
# Defensive: if upload_folder/upload_large_folder signature mismatch happens
|
| 236 |
+
print()
|
| 237 |
+
print("ERROR: Upload function raised a TypeError (likely a signature mismatch):")
|
| 238 |
+
print(f" {exc!r}")
|
| 239 |
+
print("Falling back to per-file upload...")
|
| 240 |
+
# fall through to per-file upload
|
| 241 |
+
except Exception:
|
| 242 |
+
# If upload_large_folder raised but we want to fallback to per-file, continue below.
|
| 243 |
+
pass
|
| 244 |
+
|
| 245 |
+
# If we reached here and no success message printed above, ensure the repo has content.
|
| 246 |
+
# If upload_folder/upload_large_folder raised or failed, perform per-file upload as fallback.
|
| 247 |
+
# Check whether we should perform per-file upload by seeing if repo appears empty locally (simpler: always attempt fallback if upload methods failed)
|
| 248 |
+
# Per-file upload (slower but avoids large single requests)
|
| 249 |
+
if files_to_upload:
|
| 250 |
+
# Check whether repo already has been uploaded by previous step by trying to fetch remote files?
|
| 251 |
+
# Simpler heuristic: attempt per-file upload only if previous methods likely did not raise SystemExit.
|
| 252 |
+
# We'll attempt per-file upload if the repo directory on Hugging Face is probably empty due to failure above.
|
| 253 |
+
# Perform per-file upload with retries on transient errors.
|
| 254 |
+
print("Falling back to per-file upload (this is slower but more robust for flaky networks)...")
|
| 255 |
+
for file_path, rel in files_to_upload:
|
| 256 |
+
success = False
|
| 257 |
+
try:
|
| 258 |
+
api.upload_file(
|
| 259 |
+
path_or_fileobj=str(file_path),
|
| 260 |
+
path_in_repo=rel,
|
| 261 |
+
repo_id=repo_id,
|
| 262 |
+
repo_type="dataset",
|
| 263 |
+
token=token,
|
| 264 |
+
)
|
| 265 |
+
success = True
|
| 266 |
+
except httpx.ReadTimeout:
|
| 267 |
+
print()
|
| 268 |
+
print("ERROR: Per-file upload read timed out on:", rel)
|
| 269 |
+
print("You can retry this script or use `hf upload-large-folder` / `HfApi.upload_large_folder`.")
|
| 270 |
+
sys.exit(1)
|
| 271 |
+
except Exception as exc:
|
| 272 |
+
print(f"WARNING: Failed to upload {rel!s}: {exc!r}")
|
| 273 |
+
if not success:
|
| 274 |
+
print(f"Failed to upload: {rel}")
|
| 275 |
+
# Note: this fallback may not preserve folder-level commit atomicity; it's a pragmatic fallback.
|
| 276 |
+
|
| 277 |
+
dataset_url = f"https://huggingface.co/datasets/{repo_id}"
|
| 278 |
+
print()
|
| 279 |
+
print("✅ Upload complete.")
|
| 280 |
+
print(f"Dataset is now available at: {dataset_url}")
|
| 281 |
+
print()
|
| 282 |
+
print("Next steps (recommended):")
|
| 283 |
+
print(" 1. Open the dataset page above in your browser.")
|
| 284 |
+
print(" 2. Edit the Dataset Card (README.md) to:")
|
| 285 |
+
print(" - Add paper links (e.g., your QMCSoftware/LDData arXiv paper).")
|
| 286 |
+
print(" - Add a 'Citations' section.")
|
| 287 |
+
print(" - Add 'Uses' and 'Limitations' sections, similar to:")
|
| 288 |
+
print(" - facebook/omnilingual-asr-corpus")
|
| 289 |
+
print(" - nvidia/PhysicalAI-Autonomous-Vehicles")
|
| 290 |
+
print(" - moondream/refcoco-m")
|
| 291 |
+
print(" 3. Use 'Paper' / 'Dataset' linking in the Hugging Face UI to")
|
| 292 |
+
print(" attach the dataset to your paper so it shows up on the")
|
| 293 |
+
print(" paper page and in discovery views.")
|
| 294 |
+
|
| 295 |
+
|
| 296 |
+
if __name__ == "__main__":
|
| 297 |
+
main()
|