nemo_curator.stages.interleaved.lance.reader
nemo_curator.stages.interleaved.lance.reader
Lance reader for row-wise interleaved multimodal datasets.
Module Contents
Classes
Functions
Data
API
Dataclass
Bases: CompositeStage[EmptyTask, InterleavedBatch]
Partition and read a Lance dataset as row-wise interleaved batches.
fields
fragment_ids
fragments_per_partition
include_lance_metadata
max_batch_bytes
max_batch_rows
name
path
read_kwargs
Dataclass
Bases: LanceReaderStage
Read Lance fragments into validated InterleavedBatch objects.
max_batch_bytes
max_batch_rows
name