The dataset viewer is not available for this split.
Error code: StreamingRowsError
Exception: CastError
Message: Couldn't cast
_type: string
_metadata: null
_is_non_tensor: null
lev: struct<device: string, shape: list<item: int64>, dtype: string, is_nested: bool>
child 0, device: string
child 1, shape: list<item: int64>
child 0, item: int64
child 2, dtype: string
child 3, is_nested: bool
level: struct<device: string, shape: list<item: int64>, dtype: string, is_nested: bool>
child 0, device: string
child 1, shape: list<item: int64>
child 0, item: int64
child 2, dtype: string
child 3, is_nested: bool
spatial_forcings: struct<device: string, shape: list<item: int64>, dtype: string, is_nested: bool>
child 0, device: string
child 1, shape: list<item: int64>
child 0, item: int64
child 2, dtype: string
child 3, is_nested: bool
non_spatial_forcings: struct<device: string, shape: list<item: int64>, dtype: string, is_nested: bool>
child 0, device: string
child 1, shape: list<item: int64>
child 0, item: int64
child 2, dtype: string
child 3, is_nested: bool
surface: struct<device: string, shape: list<item: int64>, dtype: string, is_nested: bool>
child 0, device: string
child 1, shape: list<item: int64>
child 0, item: int64
child 2, dtype: string
child 3, is_nested: bool
shape: list<item: null>
child 0, item: null
ozone: struct<device: string, shape: list<item: int64>, dtype: string, is_nested: bool>
child 0, device: string
child 1, shape: list<item: int64>
child 0, item: int64
child 2, dtype: string
child 3, is_nested: bool
time: struct<type: string>
child 0, type: string
device: string
to
{'surface': {'device': Value('string'), 'shape': List(Value('int64')), 'dtype': Value('string'), 'is_nested': Value('bool')}, 'level': {'device': Value('string'), 'shape': List(Value('int64')), 'dtype': Value('string'), 'is_nested': Value('bool')}, 'lev': {'device': Value('string'), 'shape': List(Value('int64')), 'dtype': Value('string'), 'is_nested': Value('bool')}, 'spatial_forcings': {'device': Value('string'), 'shape': List(Value('int64')), 'dtype': Value('string'), 'is_nested': Value('bool')}, 'non_spatial_forcings': {'device': Value('string'), 'shape': List(Value('int64')), 'dtype': Value('string'), 'is_nested': Value('bool')}, 'ozone': {'device': Value('string'), 'shape': List(Value('int64')), 'dtype': Value('string'), 'is_nested': Value('bool')}, 'time': {'type': Value('string')}, 'shape': List(Value('null')), 'device': Value('string'), '_type': Value('string')}
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 2840, in __iter__
for key, example in ex_iterable:
^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2373, in __iter__
for key, pa_table in self._iter_arrow():
~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2398, 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 536, in _iter_arrow
for key, pa_table in iterator:
^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 419, 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
_type: string
_metadata: null
_is_non_tensor: null
lev: struct<device: string, shape: list<item: int64>, dtype: string, is_nested: bool>
child 0, device: string
child 1, shape: list<item: int64>
child 0, item: int64
child 2, dtype: string
child 3, is_nested: bool
level: struct<device: string, shape: list<item: int64>, dtype: string, is_nested: bool>
child 0, device: string
child 1, shape: list<item: int64>
child 0, item: int64
child 2, dtype: string
child 3, is_nested: bool
spatial_forcings: struct<device: string, shape: list<item: int64>, dtype: string, is_nested: bool>
child 0, device: string
child 1, shape: list<item: int64>
child 0, item: int64
child 2, dtype: string
child 3, is_nested: bool
non_spatial_forcings: struct<device: string, shape: list<item: int64>, dtype: string, is_nested: bool>
child 0, device: string
child 1, shape: list<item: int64>
child 0, item: int64
child 2, dtype: string
child 3, is_nested: bool
surface: struct<device: string, shape: list<item: int64>, dtype: string, is_nested: bool>
child 0, device: string
child 1, shape: list<item: int64>
child 0, item: int64
child 2, dtype: string
child 3, is_nested: bool
shape: list<item: null>
child 0, item: null
ozone: struct<device: string, shape: list<item: int64>, dtype: string, is_nested: bool>
child 0, device: string
child 1, shape: list<item: int64>
child 0, item: int64
child 2, dtype: string
child 3, is_nested: bool
time: struct<type: string>
child 0, type: string
device: string
to
{'surface': {'device': Value('string'), 'shape': List(Value('int64')), 'dtype': Value('string'), 'is_nested': Value('bool')}, 'level': {'device': Value('string'), 'shape': List(Value('int64')), 'dtype': Value('string'), 'is_nested': Value('bool')}, 'lev': {'device': Value('string'), 'shape': List(Value('int64')), 'dtype': Value('string'), 'is_nested': Value('bool')}, 'spatial_forcings': {'device': Value('string'), 'shape': List(Value('int64')), 'dtype': Value('string'), 'is_nested': Value('bool')}, 'non_spatial_forcings': {'device': Value('string'), 'shape': List(Value('int64')), 'dtype': Value('string'), 'is_nested': Value('bool')}, 'ozone': {'device': Value('string'), 'shape': List(Value('int64')), 'dtype': Value('string'), 'is_nested': Value('bool')}, 'time': {'type': Value('string')}, 'shape': List(Value('null')), 'device': Value('string'), '_type': Value('string')}
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.
ArchesClimate-SSP data
Physical data companion to the model weights in gclyne/ArchesClimate — everything a checkpoint needs at inference time besides the weights themselves.
Code: https://github.com/grahamclyne/ArchesClimate-SSP (see INFERENCE.md there for the full
recipe, README.md's Quickstart for exact setup steps).
Layout
Almost a direct mirror of the on-disk layout this code expects under cluster.data_path —
not a repackaged format. Point cluster.data_root (see configs/cluster/external.yaml in the
code repo) at wherever you download this to, and the dataloader (CMIPForecastLeadTime) reads
it close to directly.
ipsl_scenarios/ # full raw per-scenario memmaps, unmodified (~15 GB each)
r1i1p1f1_ssp434_interpolation.memmap/
r1i1p1f1_ssp534-over_interpolation.memmap/
r1i1p1f1_ssp585_interpolation.memmap/
reference_data/ # orography.pt (IPSL grid)
reference_data_canesm5_native/ # orography.pt (CanESM5-native grid)
stats/ # normalization stats -- NOT part of the data_path
# mirror, see below
The one deviation from a pure mirror: the code's own internal name for this directory is
memmap_filled_in, a legacy name from this pipeline's history that means nothing outside of
it — so it's published here as ipsl_scenarios/ instead. No manual renaming or symlinking
needed: pass module.dataset_path=ipsl_scenarios on the command line at inference time (it's a
plain Hydra config value, read as ${cluster.data_path}${module.dataset_path}) and the code
reads straight from the download.
Each <realization>_<experiment>_interpolation.memmap/ is a tensordict.TensorDict.load_memmap()
directory (meta.json + per-field .memmap files: surface, level, lev,
spatial_forcings, non_spatial_forcings, ozone, time) — the real CMIP6 target trajectory
plus the forcing trajectory used to condition a rollout, for the full length of the scenario.
Currently published: ssp434, ssp534-over, ssp585 (all r1i1p1f1, IPSL grid).
stats/
Normalization stats ({surface,level,lev,spatial_forcings,non_spatial_forcings}_{mean,std}
tensors) keyed by the norm_scheme named in a checkpoint's config.yaml. Unlike everything
else here, this is loaded from the code repo's own stats/ directory at runtime
(importlib.resources), not from cluster.data_path — download it straight there instead (see
Usage below).
Usage
Run these from inside a clone of the code repo, via uv run (uv sync doesn't put .venv/bin
on PATH). This repo is a dataset repo, not a model repo -- hf download defaults to
--type model, so pass --type dataset explicitly or it 404s. Pick a real path for the
physical data first; don't paste <placeholder> text into a command as-is, bash reads <...>
as an input redirect:
data_dir=~/ac_data # or wherever you want it -- just not inside the code repo clone
mkdir -p "$data_dir"
uv run hf download gclyne/ArchesClimate-data --type dataset --exclude "stats/*" --local-dir "$data_dir"
uv run hf download gclyne/ArchesClimate-data --type dataset --include "stats/*" --local-dir .
(checkpoints are a separate download, from gclyne/ArchesClimate)
Then pass cluster=external cluster.data_root="$data_dir" on the command line when running the
code repo's rollout/visualization scripts — no config file to copy or edit, see its README.md
Quickstart for the full sequence.
- Downloads last month
- 95