ray.data.read_webdataset#

ray.data.read_webdataset(paths: str | List[str], *, filesystem: pyarrow.fs.FileSystem | None = None, parallelism: int = -1, arrow_open_stream_args: Dict[str, Any] | None = None, meta_provider: BaseFileMetadataProvider | None = None, partition_filter: PathPartitionFilter | None = None, decoder: bool | str | callable | list | None = True, fileselect: list | callable | None = None, filerename: list | callable | None = None, suffixes: list | callable | None = None, verbose_open: bool = False, shuffle: Literal['files'] | FileShuffleConfig | None = None, include_paths: bool = False, file_extensions: List[str] | None = None, concurrency: int | None = None, override_num_blocks: int | None = None, expand_json: bool = False) → Dataset[source]#

Create a Dataset from WebDataset files.

Parameters:

  • paths – A single file/directory path or a list of file/directory paths. A list of paths can contain both files and directories.

  • filesystem – The filesystem implementation to read from.

  • parallelism – This argument is deprecated. Use override_num_blocks argument.

  • arrow_open_stream_args – Key-word arguments passed to pyarrow.fs.FileSystem.open_input_stream. To read a compressed TFRecord file, pass the corresponding compression type (e.g. for GZIP or ZLIB, use arrow_open_stream_args={'compression': 'gzip'}).

  • meta_provider – File metadata provider. Custom metadata providers may be able to resolve file metadata more quickly and/or accurately. If None, this function uses a system-chosen implementation.

  • partition_filter – Path-based partition filter, if any. Can be used with a custom callback to read only selected partitions of a dataset.

  • decoder – A function or list of functions to decode the data.

  • fileselect – A callable or list of glob patterns to select files.

  • filerename – A function or list of tuples to rename files prior to grouping.

  • suffixes – A function or list of suffixes to select for creating samples.

  • verbose_open – Whether to print the file names as they are opened.

  • shuffle – If setting to “files”, randomly shuffle input files order before read. if setting to FileShuffleConfig, the random seed can be passed toshuffle the input files, i.e. FileShuffleConfig(seed = 42). Defaults to not shuffle with None.

  • include_paths – If True, include the path to each file. File paths are stored in the 'path' column.

  • file_extensions – A list of file extensions to filter files by.

  • concurrency – The maximum number of Ray tasks to run concurrently. Set this to control number of tasks to run concurrently. This doesn’t change the total number of tasks run or the total number of output blocks. By default, concurrency is dynamically decided based on the available resources.

  • override_num_blocks – Override the number of output blocks from all read tasks. By default, the number of output blocks is dynamically decided based on input data size and available resources. You shouldn’t manually set this value in most cases.

  • expand_json – If True, expand JSON objects into individual samples. Defaults to False.

Returns:

A Dataset that contains the example features.

Raises:

ValueError – If a file contains a message that isn’t a tf.train.Example.

PublicAPI (alpha): This API is in alpha and may change before becoming stable.