Dataset Viewer
Duplicate
The dataset viewer is not available for this split.
Cannot load the dataset split (in streaming mode) to extract the first rows.
Error code:   StreamingRowsError
Exception:    CastError
Message:      Couldn't cast
uid: string
category: string
source: string
shard: string
row: int64
kept_records: int64
selected_source_indices_sha256: string
source_records: int64
kept_indices: list<item: int64>
  child 0, item: int64
kept_ids: list<item: string>
  child 0, item: string
upstream: string
to
{'upstream': Value('string'), 'source_records': Value('int64'), 'kept_records': Value('int64'), 'selected_source_indices_sha256': Value('string'), 'kept_indices': List(Value('int64')), 'kept_ids': List(Value('string'))}
because column names don't match
Traceback:    Traceback (most recent call last):
                File "/src/services/worker/src/worker/utils.py", line 149, in get_rows_or_raise
                  return get_rows(
                      dataset=dataset,
                  ...<4 lines>...
                      column_names=column_names,
                  )
                File "/src/libs/libcommon/src/libcommon/utils.py", line 274, in decorator
                  return func(*args, **kwargs)
                File "/src/services/worker/src/worker/utils.py", line 129, in get_rows
                  rows_plus_one = list(itertools.islice(safe_iter(ds, dataset=dataset), rows_max_number + 1))
                File "/src/services/worker/src/worker/utils.py", line 544, in safe_iter
                  yield from ds.decode(False) if ds.features else ds
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 3074, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2578, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2603, in _iter_arrow
                  for key, pa_table in self.ex_iterable._iter_arrow():
                                       ~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 577, in _iter_arrow
                  for key, pa_table in iterator:
                                       ^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 459, in _iter_arrow
                  for key, pa_table in self.generate_tables_fn(**gen_kwags):
                                       ~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 373, in _generate_tables
                  self._cast_table(pa_table, json_field_paths=json_field_paths),
                  ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
                  pa_table = table_cast(pa_table, self.info.features.arrow_schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2394, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2322, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              uid: string
              category: string
              source: string
              shard: string
              row: int64
              kept_records: int64
              selected_source_indices_sha256: string
              source_records: int64
              kept_indices: list<item: int64>
                child 0, item: int64
              kept_ids: list<item: string>
                child 0, item: string
              upstream: string
              to
              {'upstream': Value('string'), 'source_records': Value('int64'), 'kept_records': Value('int64'), 'selected_source_indices_sha256': Value('string'), 'kept_indices': List(Value('int64')), 'kept_ids': List(Value('string'))}
              because column names don't match

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

GAR-v2 & GAR-Lite Data Recipe

This repository releases the lightweight audited keep-indices, reformatted annotations, and sampling indices required to reproduce the training corpora of GAR-v2 and GAR-Lite.

All training data are audited, reformatted, or sampled subsets of publicly available upstream datasets (GAR-v1, DAM-1.5M, ImageNet-21K, VCR, Osprey-QA, and Honey-Data-15M). No third-party images are redistributed here.

Contents

GAR-v2-Data-Recipe/
├── audit_indices/
│   ├── dam_kept_indices.json                  # 861,955 / 1,456,787 kept (Qwen3.5-27B multi-view audit)
│   └── imagenet_kept_indices.json             # 279,534 / 424,508 kept (Qwen3.5-27B multi-view audit)
├── osprey_gar/
│   ├── osprey_short_form_llm_gar.json         # 27,454 rows rewritten to GAR <PromptX> format
│   ├── osprey_part_level_llm_gar.json         # 37,942 rows rewritten to GAR <PromptX> format
│   ├── osprey_conversation_llm_gar.json       # 29,919 rows rewritten to GAR <PromptX> format
│   └── osprey_lvis_positive_negative_gar.json # 20,008 rows converted to GAR <PromptX> format
└── opd_90k/
    ├── region_60k_indices.jsonl               # 60,000 regional prompt indices (caption_dam / relation / mcq / mirror_qa)
    ├── honey_general_30k_indices.jsonl        # 30,000 general prompt indices into Honey-Data-15M
    └── honey_general_30k_quotas.json          # Bucket quotas and sampling seed (seed=42)

Quick Usage

See datasets.md in the main codebase for full upstream download instructions and directory layouts.

Reconstructing Audited Caption Subsets

python train/sft/data/apply_audit_indices.py \
  --source /path/to/DAM_1.5M_dataset_new.json \
  --indices data/GAR-v2-Recipe/audit_indices/dam_kept_indices.json \
  --output data/GAR-v2-Data/dam_audited_862k.json \
  --image-root /path/to/describe-anything-dataset

python train/sft/data/apply_audit_indices.py \
  --source /path/to/imagenet-21k_424508.json \
  --indices data/GAR-v2-Recipe/audit_indices/imagenet_kept_indices.json \
  --output data/GAR-v2-Data/fine_grained_audited_280k.json \
  --image-root /path/to/ImageNet-21k

Upstream Attribution & Licensing

All indices and reformatted annotations remain subject to the licenses and terms of use of their respective upstream datasets (Describe-Anything Dataset, ImageNet-21K, COCO, VCR, Osprey-724K, RGBD-Mirror, GAR-v1, and Honey-Data-15M). Intended for research use.

Downloads last month
26

Collection including laolida-w/GAR-v2-Data