File size: 4,748 Bytes
c437348 |
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 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 |
from huggingface_hub import HfApi, CommitScheduler
from typing import Any, Dict, List, Optional, Union
import uuid
from pathlib import Path
import json
import tempfile
import pyarrow as pa
import pyarrow.parquet as pq
# Initialize the ParquetScheduler
class ParquetScheduler(CommitScheduler):
"""
Usage: configure the scheduler with a repo id. Once started, you can add data to be uploaded to the Hub. 1 `.append`
call will result in 1 row in your final dataset.
```py
# Start scheduler
>>> scheduler = ParquetScheduler(repo_id="my-parquet-dataset")
# Append some data to be uploaded
>>> scheduler.append({...})
>>> scheduler.append({...})
>>> scheduler.append({...})
```
The scheduler will automatically infer the schema from the data it pushes.
Optionally, you can manually set the schema yourself:
```py
>>> scheduler = ParquetScheduler(
... repo_id="my-parquet-dataset",
... schema={
... "prompt": {"_type": "Value", "dtype": "string"},
... "negative_prompt": {"_type": "Value", "dtype": "string"},
... "guidance_scale": {"_type": "Value", "dtype": "int64"},
... "image": {"_type": "Image"},
... },
... )
See https://huggingface.co/docs/datasets/main/en/package_reference/main_classes#datasets.Value for the list of
possible values.
"""
def __init__(
self,
*,
repo_id: str,
schema: Optional[Dict[str, Dict[str, str]]] = None,
every: Union[int, float] = 5,
path_in_repo: Optional[str] = "data",
repo_type: Optional[str] = "dataset",
revision: Optional[str] = None,
private: bool = False,
token: Optional[str] = None,
allow_patterns: Union[List[str], str, None] = None,
ignore_patterns: Union[List[str], str, None] = None,
hf_api: Optional[HfApi] = None,
) -> None:
super().__init__(
repo_id=repo_id,
folder_path="dummy", # not used by the scheduler
every=every,
path_in_repo=path_in_repo,
repo_type=repo_type,
revision=revision,
private=private,
token=token,
allow_patterns=allow_patterns,
ignore_patterns=ignore_patterns,
hf_api=hf_api,
)
self._rows: List[Dict[str, Any]] = []
self._schema = schema
def append(self, row: Dict[str, Any]) -> None:
"""Add a new item to be uploaded."""
with self.lock:
self._rows.append(row)
def push_to_hub(self):
# Check for new rows to push
with self.lock:
rows = self._rows
self._rows = []
if not rows:
return
print(f"Got {len(rows)} item(s) to commit.")
# Load images + create 'features' config for datasets library
schema: Dict[str, Dict] = self._schema or {}
path_to_cleanup: List[Path] = []
for row in rows:
for key, value in row.items():
# Infer schema (for `datasets` library)
if key not in schema:
schema[key] = _infer_schema(key, value)
# Load binary files if necessary
if schema[key]["_type"] in ("Image", "Audio"):
# It's an image or audio: we load the bytes and remember to cleanup the file
file_path = Path(value)
if file_path.is_file():
row[key] = {
"path": file_path.name,
"bytes": file_path.read_bytes(),
}
path_to_cleanup.append(file_path)
# Complete rows if needed
for row in rows:
for feature in schema:
if feature not in row:
row[feature] = None
# Export items to Arrow format
table = pa.Table.from_pylist(rows)
# Add metadata (used by datasets library)
table = table.replace_schema_metadata(
{"huggingface": json.dumps({"info": {"features": schema}})}
)
# Write to parquet file
archive_file = tempfile.NamedTemporaryFile()
pq.write_table(table, archive_file.name)
# Upload
self.api.upload_file(
repo_id=self.repo_id,
repo_type=self.repo_type,
revision=self.revision,
path_in_repo=f"{uuid.uuid4()}.parquet",
path_or_fileobj=archive_file.name,
)
print("Commit completed.")
# Cleanup
archive_file.close()
for path in path_to_cleanup:
path.unlink(missing_ok=True)
|