ViperEkura
a74e5b91a3
- Store gains fetch_record/num_records alongside stream fetch/__len__ - save_bin/load_bin support per-record offsets via record_keys param - H5Store/MmapStore/JsonlStore all support dual stream+record access - DPODataset/GRPODataset use fetch_record, no cross-record concat - dpo_collate_fn + collate_fn wired through TrainConfig - fixes attention context leakage in DPO from windowed concatenation
40 lines
668 B
Python
40 lines
668 B
Python
from astrai.dataset.dataset import (
|
|
BaseDataset,
|
|
DatasetFactory,
|
|
dpo_collate_fn,
|
|
grpo_collate_fn,
|
|
)
|
|
from astrai.dataset.sampler import RDSampler
|
|
from astrai.dataset.storage import (
|
|
H5Store,
|
|
JsonlStore,
|
|
MmapStore,
|
|
Store,
|
|
StoreFactory,
|
|
detect_format,
|
|
)
|
|
from astrai.serialization import (
|
|
load_bin,
|
|
load_h5,
|
|
save_bin,
|
|
save_h5,
|
|
)
|
|
|
|
__all__ = [
|
|
"BaseDataset",
|
|
"DatasetFactory",
|
|
"dpo_collate_fn",
|
|
"grpo_collate_fn",
|
|
"Store",
|
|
"StoreFactory",
|
|
"H5Store",
|
|
"MmapStore",
|
|
"JsonlStore",
|
|
"detect_format",
|
|
"save_h5",
|
|
"load_h5",
|
|
"save_bin",
|
|
"load_bin",
|
|
"RDSampler",
|
|
]
|