Dataset Viewer
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
archive: struct<filename: string, sha256: string, size: int64>
child 0, filename: string
child 1, sha256: string
child 2, size: int64
bundles: list<item: struct<assignment_id: string, codebook_size: int64, geometry_origin: string, n_items: int (... 95 chars omitted)
child 0, item: struct<assignment_id: string, codebook_size: int64, geometry_origin: string, n_items: int64, native_ (... 83 chars omitted)
child 0, assignment_id: string
child 1, codebook_size: int64
child 2, geometry_origin: string
child 3, n_items: int64
child 4, native_index_sha256: string
child 5, num_levels: int64
child 6, tokenizer_method: string
child 7, variant: string
dataset: string
file_count: int64
files: list<item: struct<path: string, sha256: string, size: int64>>
child 0, item: struct<path: string, sha256: string, size: int64>
child 0, path: string
child 1, sha256: string
child 2, size: int64
schema: string
total_bytes: int64
native_quantizer_state_included: bool
new_run_split_protocol: string
source_split_protocol: string
category: string
predtypes: list<item: string>
child 0, item: string
to
{'archive': {'filename': Value('string'), 'sha256': Value('string'), 'size': Value('int64')}, 'category': Value('string'), 'dataset': Value('string'), 'file_count': Value('int64'), 'files': List({'path': Value('string'), 'sha256': Value('string'), 'size': Value('int64')}), 'native_quantizer_state_included': Value('bool'), 'new_run_split_protocol': Value('string'), 'predtypes': List(Value('string')), 'schema': Value('string'), 'source_split_protocol': Value('string'), 'total_bytes': Value('int64')}
because column names don't match
Traceback: Traceback (most recent call last):
File "/src/services/worker/src/worker/utils.py", line 147, 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 272, in decorator
return func(*args, **kwargs)
File "/src/services/worker/src/worker/utils.py", line 127, 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 483, 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 2951, in __iter__
for key, example in ex_iterable:
^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2461, in __iter__
for key, pa_table in self._iter_arrow():
~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2486, 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 547, in _iter_arrow
for key, pa_table in iterator:
^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 430, 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 343, 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 2378, in table_cast
return cast_table_to_schema(table, schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2306, in cast_table_to_schema
raise CastError(
...<3 lines>...
)
datasets.table.CastError: Couldn't cast
archive: struct<filename: string, sha256: string, size: int64>
child 0, filename: string
child 1, sha256: string
child 2, size: int64
bundles: list<item: struct<assignment_id: string, codebook_size: int64, geometry_origin: string, n_items: int (... 95 chars omitted)
child 0, item: struct<assignment_id: string, codebook_size: int64, geometry_origin: string, n_items: int64, native_ (... 83 chars omitted)
child 0, assignment_id: string
child 1, codebook_size: int64
child 2, geometry_origin: string
child 3, n_items: int64
child 4, native_index_sha256: string
child 5, num_levels: int64
child 6, tokenizer_method: string
child 7, variant: string
dataset: string
file_count: int64
files: list<item: struct<path: string, sha256: string, size: int64>>
child 0, item: struct<path: string, sha256: string, size: int64>
child 0, path: string
child 1, sha256: string
child 2, size: int64
schema: string
total_bytes: int64
native_quantizer_state_included: bool
new_run_split_protocol: string
source_split_protocol: string
category: string
predtypes: list<item: string>
child 0, item: string
to
{'archive': {'filename': Value('string'), 'sha256': Value('string'), 'size': Value('int64')}, 'category': Value('string'), 'dataset': Value('string'), 'file_count': Value('int64'), 'files': List({'path': Value('string'), 'sha256': Value('string'), 'size': Value('int64')}), 'native_quantizer_state_included': Value('bool'), 'new_run_split_protocol': Value('string'), 'predtypes': List(Value('string')), 'schema': Value('string'), 'source_split_protocol': Value('string'), 'total_bytes': Value('int64')}
because column names don't matchNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
OneDiffRec handoff
This repository holds the inputs for OneDiffRec experiments: the Industrial
and Office datasets, plus every Zero-Collision Reassignment (ZCR) codebook we
have. A download of this repository is a ready handoff folder that
scripts/handoff/setup.sh reads. Code:
git clone --branch leo-dev-changes https://github.com/niandd/OneDiffRec.git.
| Folder | Size | Contents |
|---|---|---|
industrial-bundle-v1/ |
291 MB | Industrial dataset: 3,105 items, native SID tables for all 27 configurations, CSVs, metadata, embeddings |
office-bundle-v1/ |
1.37 GB | Office dataset: 17,696 items, same layout |
zcr-industrial/ |
394 MB | ZCR codebooks for 8 Industrial RQ-KMeans configurations |
zcr-office/ |
1.74 GB | ZCR codebooks for all 9 Office RQ-KMeans configurations |
Each folder holds one archive, its manifest.json (every file's size and
SHA-256), SHA256SUMS and a short README. Setup verifies all of them.
ZCR codebook coverage
ZCR gives every item a unique SID by changing only the last code of items that share one. Each codebook bundle holds the ZCR SID table, the native table it came from, code arrays, the reconstructed codebooks and FAISS quantizer, and final-stage geometry. Each one was checked to re-encode every native SID.
| Dataset | RQ-KMeans | RQ-VAE | MQ |
|---|---|---|---|
| Industrial | 8 of 9 (all but 5×512) | none | none |
| Office | 9 of 9 | none | none |
The RQ-VAE and MQ quantizer checkpoints were not saved, and Industrial 5×512 was cluster-balanced, so those configurations have no ZCR codebook yet. Training skips them and lists them; native-SID runs cover all 27 configurations.
Steps (Slurm login node, inside the code checkout)
# 1. Get the files: everything from here (3.8 GB) ...
bash scripts/handoff/download.sh /path/to/onediffrec-handoff-v1
# ... or, if you received onediffrec-handoff-v1.zip, unzip it and add the codebooks it lacks (2.1 GB):
# bash scripts/handoff/download.sh /path/to/onediffrec-handoff-v1 --codebooks-only
# 2. Setup, once (about 20 minutes)
bash scripts/handoff/setup.sh /path/to/onediffrec-handoff-v1
# 3. Train and evaluate (each command submits Slurm jobs and returns)
bash scripts/handoff/run_zcr.sh # ZCR: Industrial 8 + Office 9 configurations, 68 runs
bash scripts/handoff/run_office_next2.sh # Office native SIDs, next-two-item, 54 runs
# 4. Progress, and one results CSV per dataset in work/results/
bash scripts/handoff/results.sh
download.sh needs a Hugging Face token that can read this repository: export
HF_TOKEN, or paste the token when it asks. Add --smoke to a run command for
a 15-minute check first. The repository README has the details.
Adding codebooks later
Setup extracts zcr-DATASET/ to work/DATASET/zcr/DATASET/METHOD_Dcodebook_S/,
and training reads that tree. To add codebooks for more configurations:
- A new package (
zcr-DATASET/fromscripts/data_bundle.py pack-zcr): replace the folder in the handoff folder and rerunsetup.sh. - A single bundle folder: copy it to
work/DATASET/zcr/DATASET/METHOD_Dcodebook_S/.
Then train only the new configurations under a new run name, e.g.
RUN_NAME=zcr-v2 bash scripts/handoff/run_zcr.sh --dataset office --method rqvae.
Terms
These are derived Amazon review data. This card grants no new license to the source item and review content.
- Downloads last month
- 4