WeHub Mirror
61 lines
1.9 KiB
Python
61 lines
1.9 KiB
Python
import warnings
|
|
from io import BytesIO
|
|
from typing import TYPE_CHECKING, Any, Dict, Iterator, List, Optional, Union
|
|
|
|
import numpy as np
|
|
|
|
from ray.data.block import Block, BlockAccessor
|
|
from ray.data.datasource.file_based_datasource import FileBasedDatasource
|
|
|
|
if TYPE_CHECKING:
|
|
import pyarrow
|
|
|
|
|
|
class NumpyDatasource(FileBasedDatasource):
|
|
"""Numpy datasource, for reading and writing Numpy files."""
|
|
|
|
_COLUMN_NAME = "data"
|
|
_FILE_EXTENSIONS = ["npy"]
|
|
|
|
def __init__(
|
|
self,
|
|
paths: Union[str, List[str]],
|
|
numpy_load_args: Optional[Dict[str, Any]] = None,
|
|
allow_pickle: bool = False,
|
|
**file_based_datasource_kwargs,
|
|
):
|
|
super().__init__(paths, **file_based_datasource_kwargs)
|
|
|
|
if numpy_load_args is None:
|
|
numpy_load_args = {}
|
|
|
|
if "allow_pickle" in numpy_load_args:
|
|
warnings.warn(
|
|
"`allow_pickle` in `numpy_load_args` is ignored. "
|
|
"Use `read_numpy(..., allow_pickle=True)` instead "
|
|
"if you want to load object-dtype .npy files.",
|
|
stacklevel=2,
|
|
)
|
|
numpy_load_args = {
|
|
k: v for k, v in numpy_load_args.items() if k != "allow_pickle"
|
|
}
|
|
self.allow_pickle = allow_pickle
|
|
self.numpy_load_args = numpy_load_args
|
|
|
|
def _read_stream(self, f: "pyarrow.NativeFile", path: str) -> Iterator[Block]:
|
|
# TODO(ekl) Ideally numpy can read directly from the file, but it
|
|
# seems like it requires the file to be seekable.
|
|
buf = BytesIO()
|
|
data = f.readall()
|
|
buf.write(data)
|
|
buf.seek(0)
|
|
yield BlockAccessor.batch_to_block(
|
|
{
|
|
"data": np.load(
|
|
buf,
|
|
allow_pickle=self.allow_pickle,
|
|
**self.numpy_load_args,
|
|
)
|
|
}
|
|
)
|