Continual Learning with Soft Prompt
Collection
Private reproducibility artifacts for distilling a mathematical reasoning RL checkpoint into a 16-token soft prompt. • 5 items • Updated
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
update: int64
loss: double
response_tokens: int64
gradient_norm: double
learning_rate: double
effective_batch: int64
loss_normalization: string
gradient_norm_after_clip: double
seconds: double
elapsed_seconds: double
global_batch_sha256: string
gpu_max_memory_bytes: int64
schema: string
selection_candidate: bool
bytes: int64
sha256: string
to
{'bytes': Value('int64'), 'schema': Value('string'), 'selection_candidate': Value('bool'), 'sha256': Value('string'), 'update': Value('int64')}
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
update: int64
loss: double
response_tokens: int64
gradient_norm: double
learning_rate: double
effective_batch: int64
loss_normalization: string
gradient_norm_after_clip: double
seconds: double
elapsed_seconds: double
global_batch_sha256: string
gpu_max_memory_bytes: int64
schema: string
selection_candidate: bool
bytes: int64
sha256: string
to
{'bytes': Value('int64'), 'schema': Value('string'), 'selection_candidate': Value('bool'), 'sha256': Value('string'), 'update': 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.
Continual Learning with Soft Prompt — public experiment archive
Five-domain source data, teacher generations, grading/selection records, available packed training sets, runtime source files and selected native checkpoints for Issue #18.
Directories: math/, code/, science/, if/, agent/, runtime/, checkpoints/. See MANIFEST.json for exact bytes and SHA256 and MIGRATION-SCOPE.json for provenance and remaining work.
Included checkpoints
Original Math180, Code140, Science160, IF100, Agent140, plus frequency-initialized Math180. Full native checkpoint files are preserved. Base/teacher weights are referenced by fixed Hugging Face revision and are not included.
Readiness and limitations
Math/Code/Science packed data are included. IF training provenance identifies 14,110 records; Agent cap4 identifies 3,912 trajectories. Their exact packed token files still require acquisition or verified reconstruction. Raw return packages are not interchangeable with these packed training sets.
The initial360-file release excluded new Math. The verified training supplement below now includes the accepted cap1 set and packed training data; detailed generation/evaluation raw remains excluded from the supplement.
This archive is not yet a one-command portable execution environment. Source path mapping, dependency reconstruction and fresh-server smoke tests remain required. Resume generation only after verifying request identities, completed responses and exclusive execution ownership; never blindly restart the whole source plan. Original archive members may include local source paths and sealed ledgers.
Teacher-generation randomness in the older fast runs is engine-seeded, not independently seeded per request; exact token replay is not promised. Changing accelerator/backend requires a new compatibility check.
Native state.pt files use PyTorch serialization. Only load trusted artifacts and verify SHA256 first. The owner explicitly authorized public publication of this reviewed archive. Only this unified destination repository is made public; original source repositories are unchanged. The historical MIGRATION-SCOPE.json describes the initial private upload proposal, superseded by this public authorization.
Latest Math and training backup supplement
The owner authorized supplements/20261105-training-backup/: latest11508-question cap1 Math accepted candidates, exact packed training data, generation plans, current source/config snapshot, and all included committed prompt-training states. See its MANIFEST.json for exact contents. No detailed benchmark response archives are added by this supplement. IF/Agent original generation packages remain in the initial release; their exact final packed token files are still a documented migration gap.
- Downloads last month
- 2,101