The dataset viewer is not available for this split.
Error code: StreamingRowsError
Exception: TypeError
Message: Couldn't cast array of type
struct<allen_model_id: int64, allen_specimen_id: int64, axon: string, backend: string, biophysics_hoc_sha256: null, celsius: double, code_revision: struct<commit: string, dirty: bool>, data_version: string, dt_ms: double, exc_gmax_scale: double, fit_parameters_sha256: string, inh_gmax_scale: double, input_routing: string, input_sampler: string, integrator: string, junction_potential_mv: double, mechanism_files: list<item: string>, mechanisms_sha256: string, morphology_sha256: string, neuron_version: string, nseg_policy: string, soma_samples_per_ms: int64, spike_label_rule: string, synapse_delay_ms: double, synapse_objects: string, synapses: string, teacher: string, template_hoc_sha256: null, v_init_mv: double>
to
{'backend': Value('string'), 'biophysics_hoc_sha256': Value('string'), 'code_revision': {'commit': Value('string'), 'dirty': Value('bool')}, 'data_version': Value('string'), 'dt_ms': Value('float64'), 'exc_gmax_scale': Value('float64'), 'inh_gmax_scale': Value('float64'), 'input_routing': Value('string'), 'input_sampler': Value('string'), 'integrator': Value('string'), 'mechanism_files': List(Value('string')), 'mechanisms_sha256': Value('string'), 'morphology_sha256': Value('string'), 'neuron_version': Value('string'), 'nseg_policy': Value('string'), 'soma_samples_per_ms': Value('int64'), 'spike_label_rule': Value('string'), 'synapse_delay_ms': Value('float64'), 'synapses': Value('string'), 'teacher': Value('string'), 'template_hoc_sha256': Value('string'), 'v_init_mv': Value('float64')}
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 2312, in cast_table_to_schema
cast_array_to_feature(
~~~~~~~~~~~~~~~~~~~~~^
table[name] if name in table_column_names else pa.array([None] * len(table), type=schema.field(name).type),
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
feature,
^^^^^^^^
)
^
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 1861, in wrapper
return pa.chunked_array([func(chunk, *args, **kwargs) for chunk in array.chunks])
~~~~^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2158, in cast_array_to_feature
raise TypeError(f"Couldn't cast array of type\n{_short_str(array.type)}\nto\n{_short_str(feature)}")
TypeError: Couldn't cast array of type
struct<allen_model_id: int64, allen_specimen_id: int64, axon: string, backend: string, biophysics_hoc_sha256: null, celsius: double, code_revision: struct<commit: string, dirty: bool>, data_version: string, dt_ms: double, exc_gmax_scale: double, fit_parameters_sha256: string, inh_gmax_scale: double, input_routing: string, input_sampler: string, integrator: string, junction_potential_mv: double, mechanism_files: list<item: string>, mechanisms_sha256: string, morphology_sha256: string, neuron_version: string, nseg_policy: string, soma_samples_per_ms: int64, spike_label_rule: string, synapse_delay_ms: double, synapse_objects: string, synapses: string, teacher: string, template_hoc_sha256: null, v_init_mv: double>
to
{'backend': Value('string'), 'biophysics_hoc_sha256': Value('string'), 'code_revision': {'commit': Value('string'), 'dirty': Value('bool')}, 'data_version': Value('string'), 'dt_ms': Value('float64'), 'exc_gmax_scale': Value('float64'), 'inh_gmax_scale': Value('float64'), 'input_routing': Value('string'), 'input_sampler': Value('string'), 'integrator': Value('string'), 'mechanism_files': List(Value('string')), 'mechanisms_sha256': Value('string'), 'morphology_sha256': Value('string'), 'neuron_version': Value('string'), 'nseg_policy': Value('string'), 'soma_samples_per_ms': Value('int64'), 'spike_label_rule': Value('string'), 'synapse_delay_ms': Value('float64'), 'synapses': Value('string'), 'teacher': Value('string'), 'template_hoc_sha256': Value('string'), 'v_init_mv': Value('float64')}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.
AxoBench Population
AxoBench Population contains population-context neuron traces for training and evaluating learned neuron surrogates under connected-network input distributions. Connected AxoSim rollouts supply the input histories; role-correct biophysical teachers supply the target voltage and spike traces.
This is the AxoBench v2 version. It replays the exact input windows of the v1
release (population/ in Axym-Labs/axobench-v1)
through the v2 teachers and simulator settings of
Axym-Labs/axobench.
Size and splits
| Split | Contexts | Traces | NPZ shards |
|---|---|---|---|
train |
80 | 40,000 | 1,040 |
val |
20 | 10,000 | 260 |
| total | 100 | 50,000 | 1,300 |
Contexts, rather than neighboring windows from one context, define the train/validation boundary. Each context contributes five 1,200-ms windows for each of its 100 target neurons, with exact 80/20 excitatory/inhibitory composition.
Teachers
- Excitatory targets: the 12 training cells of AxoBench v2 (11 Allen Cell Types all-active layer-5 models and the NeuronIO layer-5b cell), assigned round-robin within each context. The three held-out v2 cells are not used, so the zero-shot track stays unseen.
- Inhibitory targets: the Allen all-active Sst layer-5 model of specimen 482690728, with the excitatory efficacy fitted for v1 (0.70); under the v2 settings it reproduces the v1 teacher's firing rate on the checked windows.
Simulator settings follow v2: synapse delay 0 ms, all 1,278 input sites connected, d_lambda compartments, per-cell calibrated excitatory gains, and the somatic voltage stored at 4 kHz in addition to 1 ms.
Shard schema
Each compressed NPZ shard holds one teacher on one context:
inputs:int8[trace, 1200, 1278], the v1 input windows;targets:float32[trace, 1200, 2], spike indicator and (min(V, -55 mV) + 67.7 mV) / 10;soma_mv:float32[trace, 1200]andsoma_mv_hr:float32[trace, 4800];sample_ids,target_morphology,target_role,target_is_inhibitory,split;context_index,context_iteration,context_seed,context_kind,target_population_indexandwindow_start_msfrom v1.
Each shard has a JSON sidecar with the complete generator settings.
train/manifest.json and val/manifest.json list shard counts, sample IDs and
SHA-256 hashes; package-manifest.json summarizes the release and
checksums.sha256 covers every file.
Validation boundary
All shards passed the release validator for schema, finite values, binary spike labels, sample counts, unique sample IDs, exact role balance, manifest hashes, and disjoint context splits. This establishes package integrity and provenance; it does not imply that one fixed training recipe will improve every connected population metric.
Loading
Download with axobench download-data --population or
huggingface_hub.snapshot_download, then load a shard with numpy.load.
from pathlib import Path
import numpy as np
shard = next(Path("train").glob("*.npz"))
with np.load(shard, allow_pickle=False) as data:
inputs = data["inputs"]
targets = data["targets"]
License and attribution
The data are released under CC BY-NC 4.0. The teachers are simulated on models from the Allen Cell Types Database and on the NeuronIO layer-5b cell (Hay et al. 2011; Beniaguev et al. 2021); check the terms of these resources for your application.
- Downloads last month
- -