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
uid: string
split: string
index: int64
seconds: double
source_ids: list<item: string>
child 0, item: string
speakers: list<item: string>
child 0, item: string
corpora: list<item: string>
child 0, item: string
whisper: list<item: bool>
child 0, item: bool
transcripts: list<item: string>
child 0, item: string
source_samples: list<item: int64>
child 0, item: int64
offsets_samples: list<item: int64>
child 0, item: int64
schedule: struct<mode: string, requested: double, actual_energy_proxy: double, requested_mode_unavailable: str (... 4 chars omitted)
child 0, mode: string
child 1, requested: double
child 2, actual_energy_proxy: double
child 3, requested_mode_unavailable: string
circular_room_id: string
level_prior: struct<sir_db: double, snr_db: double, speaker_bucket: string, noise_bucket: string, prior_status: s (... 30 chars omitted)
child 0, sir_db: double
child 1, snr_db: double
child 2, speaker_bucket: string
child 3, noise_bucket: string
child 4, prior_status: string
child 5, snr_definition: string
microphone_filter_hz: list<item: double>
child 0, item: double
requested_recording_dbfs: double
capture: string
noise_mode: string
av_offset_seconds: double
path: string
fs: int64
fps: int64
revision: string
files: struct<mixture: struct<file: string, channels: int64, sha256: string>, source_a: struct<file: string (... 371 chars omitted)
child 0, mixture: struct<file: string, channels: int64, sha256: string>
child 0, file: string
...
311 chars omitted)
child 0, tr: struct<count: int64, selection_sha256: string, manifest_sha256: string, shards: list<item: struct<pa (... 29 chars omitted)
child 0, count: int64
child 1, selection_sha256: string
child 2, manifest_sha256: string
child 3, shards: list<item: struct<path: string, sha256: string>>
child 0, item: struct<path: string, sha256: string>
child 0, path: string
child 1, sha256: string
child 1, cv: struct<count: int64, selection_sha256: string, manifest_sha256: string, shards: list<item: struct<pa (... 29 chars omitted)
child 0, count: int64
child 1, selection_sha256: string
child 2, manifest_sha256: string
child 3, shards: list<item: struct<path: string, sha256: string>>
child 0, item: struct<path: string, sha256: string>
child 0, path: string
child 1, sha256: string
child 2, tt: struct<count: int64, selection_sha256: string, manifest_sha256: string, shards: list<item: struct<pa (... 29 chars omitted)
child 0, count: int64
child 1, selection_sha256: string
child 2, manifest_sha256: string
child 3, shards: list<item: struct<path: string, sha256: string>>
child 0, item: struct<path: string, sha256: string>
child 0, path: string
child 1, sha256: string
manifests: struct<tr: string, cv: string, tt: string>
child 0, tr: string
child 1, cv: string
child 2, tt: string
to
{'status': Value('string'), 'counts': {'tr': Value('int64'), 'cv': Value('int64'), 'tt': Value('int64')}, 'manifests': {'tr': Value('string'), 'cv': Value('string'), 'tt': Value('string')}, 'format': Value('string'), 'source_ready_sha256': Value('string'), 'splits': {'tr': {'count': Value('int64'), 'selection_sha256': Value('string'), 'manifest_sha256': Value('string'), 'shards': List({'path': Value('string'), 'sha256': Value('string')})}, 'cv': {'count': Value('int64'), 'selection_sha256': Value('string'), 'manifest_sha256': Value('string'), 'shards': List({'path': Value('string'), 'sha256': Value('string')})}, 'tt': {'count': Value('int64'), 'selection_sha256': Value('string'), 'manifest_sha256': Value('string'), 'shards': List({'path': Value('string'), 'sha256': Value('string')})}}, 'seconds': Value('float64')}
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 272, 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 3052, in __iter__
for key, example in ex_iterable:
^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2559, in __iter__
for key, pa_table in self._iter_arrow():
~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2584, 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 558, in _iter_arrow
for key, pa_table in iterator:
^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 440, 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 361, 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 2381, in table_cast
return cast_table_to_schema(table, schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2309, in cast_table_to_schema
raise CastError(
...<3 lines>...
)
datasets.table.CastError: Couldn't cast
uid: string
split: string
index: int64
seconds: double
source_ids: list<item: string>
child 0, item: string
speakers: list<item: string>
child 0, item: string
corpora: list<item: string>
child 0, item: string
whisper: list<item: bool>
child 0, item: bool
transcripts: list<item: string>
child 0, item: string
source_samples: list<item: int64>
child 0, item: int64
offsets_samples: list<item: int64>
child 0, item: int64
schedule: struct<mode: string, requested: double, actual_energy_proxy: double, requested_mode_unavailable: str (... 4 chars omitted)
child 0, mode: string
child 1, requested: double
child 2, actual_energy_proxy: double
child 3, requested_mode_unavailable: string
circular_room_id: string
level_prior: struct<sir_db: double, snr_db: double, speaker_bucket: string, noise_bucket: string, prior_status: s (... 30 chars omitted)
child 0, sir_db: double
child 1, snr_db: double
child 2, speaker_bucket: string
child 3, noise_bucket: string
child 4, prior_status: string
child 5, snr_definition: string
microphone_filter_hz: list<item: double>
child 0, item: double
requested_recording_dbfs: double
capture: string
noise_mode: string
av_offset_seconds: double
path: string
fs: int64
fps: int64
revision: string
files: struct<mixture: struct<file: string, channels: int64, sha256: string>, source_a: struct<file: string (... 371 chars omitted)
child 0, mixture: struct<file: string, channels: int64, sha256: string>
child 0, file: string
...
311 chars omitted)
child 0, tr: struct<count: int64, selection_sha256: string, manifest_sha256: string, shards: list<item: struct<pa (... 29 chars omitted)
child 0, count: int64
child 1, selection_sha256: string
child 2, manifest_sha256: string
child 3, shards: list<item: struct<path: string, sha256: string>>
child 0, item: struct<path: string, sha256: string>
child 0, path: string
child 1, sha256: string
child 1, cv: struct<count: int64, selection_sha256: string, manifest_sha256: string, shards: list<item: struct<pa (... 29 chars omitted)
child 0, count: int64
child 1, selection_sha256: string
child 2, manifest_sha256: string
child 3, shards: list<item: struct<path: string, sha256: string>>
child 0, item: struct<path: string, sha256: string>
child 0, path: string
child 1, sha256: string
child 2, tt: struct<count: int64, selection_sha256: string, manifest_sha256: string, shards: list<item: struct<pa (... 29 chars omitted)
child 0, count: int64
child 1, selection_sha256: string
child 2, manifest_sha256: string
child 3, shards: list<item: struct<path: string, sha256: string>>
child 0, item: struct<path: string, sha256: string>
child 0, path: string
child 1, sha256: string
manifests: struct<tr: string, cv: string, tt: string>
child 0, tr: string
child 1, cv: string
child 2, tt: string
to
{'status': Value('string'), 'counts': {'tr': Value('int64'), 'cv': Value('int64'), 'tt': Value('int64')}, 'manifests': {'tr': Value('string'), 'cv': Value('string'), 'tt': Value('string')}, 'format': Value('string'), 'source_ready_sha256': Value('string'), 'splits': {'tr': {'count': Value('int64'), 'selection_sha256': Value('string'), 'manifest_sha256': Value('string'), 'shards': List({'path': Value('string'), 'sha256': Value('string')})}, 'cv': {'count': Value('int64'), 'selection_sha256': Value('string'), 'manifest_sha256': Value('string'), 'shards': List({'path': Value('string'), 'sha256': Value('string')})}, 'tt': {'count': Value('int64'), 'selection_sha256': Value('string'), 'manifest_sha256': Value('string'), 'shards': List({'path': Value('string'), 'sha256': Value('string')})}}, 'seconds': Value('float64')}
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.
SAVE: paper split
This repository contains the paper's selected 35,768 synthetic scenes: 32,768 training, 1,000 validation and 2,000 test scenes. It is a subset of the larger SAVE v2 release, not an 80,000-scene training release. The training subset covers 240 speakers, 16,065 source clips and 6,000 acoustic configurations, totaling approximately 119.42 mixture-hours.
Each scene includes 16 kHz six-channel mixture audio, two source signals, direct references, two 25 fps mouth-crop sequences with validity indicators, localization labels and scene metadata. The reference microphone geometry is a six-channel circular array of radius 3.5 cm. Synthetic microphone response, timing, visual and environmental perturbations are included; they do not establish calibration to an arbitrary physical device.
Format
manifests/{tr,cv,tt}.jsonl identifies each scene by a relative shards/<split>-<number>.pack@<offset> location. At that offset, an unsigned 64-bit little-endian length precedes a ZIP_STORED scene record. record_sha256 hashes the record payload; READY.json contains split and shard hashes. original_index retains the original parent-release index. Scene media and labels are preserved byte-for-byte; the manifest locations are made relative.
The SAV-BandNet implementation includes the reader and full-training/test commands. Packed records avoid excessive small-file operations. Do not treat the .pack files as conventional ZIP archives without reading their record offsets.
Evaluation scope
The validation and test splits have been repeatedly inspected during development. They are not untouched confirmation sets. Preserve split identities and report the exact checkpoint, window configuration, waveform normalization and metric definition when comparing methods. Full-scene SI-SNR and overlap-only SI-SNR are different measurements.
Source assets and terms
The synthetic release derives from existing audiovisual speech, impulse-response and noise resources. Source metadata is retained within scene records. Synthesis and packaging do not replace the original assets' licenses or participant permissions. Retained notices and provenance accompany the repository; no single permissive license is asserted for every underlying asset, and commercial-use clearance is not established by this release.
- Downloads last month
- 427