The dataset viewer is not available for this split.
Error code: StreamingRowsError
Exception: CastError
Message: Couldn't cast
computed_utc: timestamp[s]
target: string
n_molecules: int64
seeds: int64
max_iter: int64
hts_dropped_as_leakage: list<item: string>
child 0, item: string
structural_spread: double
wall_seconds: double
rows: list<item: struct<block: string, width: int64, n: int64, spearman: double, sd: double, mae: double, (... 45 chars omitted)
child 0, item: struct<block: string, width: int64, n: int64, spearman: double, sd: double, mae: double, null: doubl (... 33 chars omitted)
child 0, block: string
child 1, width: int64
child 2, n: int64
child 3, spearman: double
child 4, sd: double
child 5, mae: double
child 6, null: double
child 7, gain: double
child 8, seconds: double
bits: int64
frac_instances_that_are_repeats: double
mean_total_count: double
mean_occupied_bits: double
additive_mean_delta: double
radius: int64
recognition_mean_delta: double
to
{'n_molecules': Value('int64'), 'radius': Value('int64'), 'bits': Value('int64'), 'seeds': List(Value('int64')), 'mean_occupied_bits': Value('float64'), 'mean_total_count': Value('float64'), 'frac_instances_that_are_repeats': Value('float64'), 'additive_mean_delta': Value('float64'), 'recognition_mean_delta': Value('float64'), 'rows': List({'target': Value('string'), 'n': Value('int64'), 'binary': Value('float64'), 'count': Value('float64'), 'delta': Value('float64'), 'binary_sd': Value('float64'), 'count_sd': Value('float64'), 'null': Value('float64')})}
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
computed_utc: timestamp[s]
target: string
n_molecules: int64
seeds: int64
max_iter: int64
hts_dropped_as_leakage: list<item: string>
child 0, item: string
structural_spread: double
wall_seconds: double
rows: list<item: struct<block: string, width: int64, n: int64, spearman: double, sd: double, mae: double, (... 45 chars omitted)
child 0, item: struct<block: string, width: int64, n: int64, spearman: double, sd: double, mae: double, null: doubl (... 33 chars omitted)
child 0, block: string
child 1, width: int64
child 2, n: int64
child 3, spearman: double
child 4, sd: double
child 5, mae: double
child 6, null: double
child 7, gain: double
child 8, seconds: double
bits: int64
frac_instances_that_are_repeats: double
mean_total_count: double
mean_occupied_bits: double
additive_mean_delta: double
radius: int64
recognition_mean_delta: double
to
{'n_molecules': Value('int64'), 'radius': Value('int64'), 'bits': Value('int64'), 'seeds': List(Value('int64')), 'mean_occupied_bits': Value('float64'), 'mean_total_count': Value('float64'), 'frac_instances_that_are_repeats': Value('float64'), 'additive_mean_delta': Value('float64'), 'recognition_mean_delta': Value('float64'), 'rows': List({'target': Value('string'), 'n': Value('int64'), 'binary': Value('float64'), 'count': Value('float64'), 'delta': Value('float64'), 'binary_sd': Value('float64'), 'count_sd': Value('float64'), 'null': 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.
CYP Fingerprint Bundle
Precomputed molecular representations for the marimo notebook "Five ways to describe a molecule, and the one that had to be computed", submitted to Bring Cheminformatics to Life: molab Notebook Competition #3 (OpenADMET x marimo).
This exists for one reason: two of the five representations cannot be computed inside a hosted notebook kernel in reasonable time. Everything cheap is rebuilt live in the notebook so a reader can change it and watch it move.
What is in it
cyp_fingerprint_bundle.parquet -- 9,154 molecules x 1,541 columns.
| block | columns | why precomputed |
|---|---|---|
L6:* quantum |
185 | GFN2-xTB SCF on 3 ETKDG+MMFF conformers, ~0.1 s/molecule |
MORD:* Mordred |
1,335 | ~0.10 s/molecule |
HTS:* |
11 | measured values from the PubChem NCGC panel |
smiles, y_*, split |
10 | identifiers, labels, split membership |
Computed live in the notebook instead: Morgan/ECFP4, and an atom-based physics block (VSA / EState / 2D autocorrelation).
cyp2c19_holdout_result.json -- the full-size held-out CYP2C19 comparison
(3,249 molecules, 3 seeds, 8 representation blocks, each against its own
shuffled-label null).
Splits, and one thing worth knowing
| split | n | with quantum |
|---|---|---|
| train | 4,905 | 4,895 (99.8%) |
| holdout (PubChem CYP2C19) | 3,499 | 3,210 (91.7%) |
| blinded test | 750 | 0 |
The xTB corpus is a training corpus, so the challenge's 750 blinded compounds were never in it. They are carried with quantum columns honestly null, and every quantum panel in the notebook drops them explicitly rather than imputing.
The 750 blinded compounds carry no target column. No blinded label is published here.
Provenance and licence
- OpenADMET CYP Inhibition Blind Challenge, training split -- CC-BY-4.0
- PubChem NCGC CYP panel, AIDs 410/883/884/885/891/899 -- public domain
- Quantum block: GFN2-xTB via
tblite; Mordred viamordred
Released CC-BY-4.0, matching the upstream challenge data.
- Downloads last month
- 46