How to use the hdf5plugin.LZ4 function in hdf5plugin

assert np.issubdtype(batch_like.dtype, np.number), "invalid non-flattened array subtype"
        auto_chunker = False
        if chunk_size is None:
            auto_chunker = True
            chunk_size = (1, *batch_like.shape[1:])
        chunk_byte_size = np.multiply.reduce(chunk_size) * batch_like.dtype.itemsize
        assert auto_chunker or 10 * (2 ** 10) <= chunk_byte_size < 2 ** 20, \
            f"unrecommended chunk byte size ({chunk_byte_size}) should be in [10KiB,1MiB];" \
            " see http://docs.h5py.org/en/stable/high/dataset.html#chunked-storage"
        if compression == "chunk_lz4":
            dset = fd.create_dataset(
                name=name,
                shape=(max_len, *batch_like.shape[1:]),
                chunks=chunk_size,
                dtype=batch_like.dtype,
                **hdf5plugin.LZ4(nbytes=0)
            )
        else:
            assert compression not in no_compression_flags or len(compression_args) == 0
            dset = fd.create_dataset(
                name=name,
                shape=(max_len, *batch_like.shape[1:]),
                chunks=chunk_size,
                dtype=batch_like.dtype,
                compression=compression if compression not in no_compression_flags else None,
                **compression_args
            )
        dset.attrs["orig_shape"] = batch_like.shape[1:]  # removes batch dim
    else:
        assert thelper.utils.is_scalar(batch_like[0])
        if np.issubdtype(batch_like.dtype, np.number):
            assert compression in no_compression_flags, "cannot compress scalar elements"