Dataset Viewer
Duplicate
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
version: string
scope: string
repo_id: string
card: string
stage1: struct<variants_per_material: int64, excitations: string, shards: int64>
  child 0, variants_per_material: int64
  child 1, excitations: string
  child 2, shards: int64
stage2_spatial_profiles: list<item: null>
  child 0, item: null
num_particles: int64
num_steps: int64
num_grids: int64
dt: double
constitutive_update_order: string
stage1_trajectories: int64
recipe: struct<version: string, scope: string, repo_id: string, card: string, stage1: struct<variants_per_ma (... 199 chars omitted)
  child 0, version: string
  child 1, scope: string
  child 2, repo_id: string
  child 3, card: string
  child 4, stage1: struct<variants_per_material: int64, excitations: string, shards: int64>
      child 0, variants_per_material: int64
      child 1, excitations: string
      child 2, shards: int64
  child 5, stage2_spatial_profiles: list<item: null>
      child 0, item: null
  child 6, num_particles: int64
  child 7, num_steps: int64
  child 8, num_grids: int64
  child 9, dt: double
  child 10, constitutive_update_order: string
stage2_trajectories: int64
archives: list<item: struct<file: string, bytes: int64, trajectories: int64>>
  child 0, item: struct<file: string, bytes: int64, trajectories: int64>
      child 0, file: string
      child 1, bytes: int64
      child 2, trajectories: int64
to
{'recipe': {'version': Value('string'), 'scope': Value('string'), 'repo_id': Value('string'), 'card': Value('string'), 'stage1': {'variants_per_material': Value('int64'), 'excitations': Value('string'), 'shards': Value('int64')}, 'stage2_spatial_profiles': List(Value('null')), 'num_particles': Value('int64'), 'num_steps': Value('int64'), 'num_grids': Value('int64'), 'dt': Value('float64'), 'constitutive_update_order': Value('string')}, 'stage1_trajectories': Value('int64'), 'stage2_trajectories': Value('int64'), 'archives': List({'file': Value('string'), 'bytes': Value('int64'), 'trajectories': Value('int64')})}
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
              version: string
              scope: string
              repo_id: string
              card: string
              stage1: struct<variants_per_material: int64, excitations: string, shards: int64>
                child 0, variants_per_material: int64
                child 1, excitations: string
                child 2, shards: int64
              stage2_spatial_profiles: list<item: null>
                child 0, item: null
              num_particles: int64
              num_steps: int64
              num_grids: int64
              dt: double
              constitutive_update_order: string
              stage1_trajectories: int64
              recipe: struct<version: string, scope: string, repo_id: string, card: string, stage1: struct<variants_per_ma (... 199 chars omitted)
                child 0, version: string
                child 1, scope: string
                child 2, repo_id: string
                child 3, card: string
                child 4, stage1: struct<variants_per_material: int64, excitations: string, shards: int64>
                    child 0, variants_per_material: int64
                    child 1, excitations: string
                    child 2, shards: int64
                child 5, stage2_spatial_profiles: list<item: null>
                    child 0, item: null
                child 6, num_particles: int64
                child 7, num_steps: int64
                child 8, num_grids: int64
                child 9, dt: double
                child 10, constitutive_update_order: string
              stage2_trajectories: int64
              archives: list<item: struct<file: string, bytes: int64, trajectories: int64>>
                child 0, item: struct<file: string, bytes: int64, trajectories: int64>
                    child 0, file: string
                    child 1, bytes: int64
                    child 2, trajectories: int64
              to
              {'recipe': {'version': Value('string'), 'scope': Value('string'), 'repo_id': Value('string'), 'card': Value('string'), 'stage1': {'variants_per_material': Value('int64'), 'excitations': Value('string'), 'shards': Value('int64')}, 'stage2_spatial_profiles': List(Value('null')), 'num_particles': Value('int64'), 'num_steps': Value('int64'), 'num_grids': Value('int64'), 'dt': Value('float64'), 'constitutive_update_order': Value('string')}, 'stage1_trajectories': Value('int64'), 'stage2_trajectories': Value('int64'), 'archives': List({'file': Value('string'), 'bytes': Value('int64'), 'trajectories': Value('int64')})}
              because column names don't match

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.

Inverse Physics Warp Dataset — historical Stage-2 targets

This initial public release contains five complete homogeneous Stage-2 targets. It contains no Stage-1 training corpus and is not the planned full v1 release.

The five materials are jelly, sand, water, plasticine, and non-Newtonian material. Each target has 10,000 particles and 2,000 Warp MPM steps, dt=1/4800 s and a 50-cell grid per axis. They share the reference-cube geometry and initial state. The update order is stress → MPM → plastic projection. The archive contains the original stored tensor values; this release does not regenerate historical GT.

Each trajectory contains positions (GtX.pt), velocities (GtV.pt), APIC C (GtC.pt), projected deformation (GtF.pt), trial deformation (GtFtmp.pt), stress (GtStress.pt), reference coordinates (CanonicalX.pt), config.yaml, and numerical validation.json. Load tensors with torch.load(..., weights_only=True). Positions and velocities have shape [2001,10000,3]; state/stress tensor shapes and the exact constitutive parameters are specified by the accompanying files.

For position-only inverse experiments, internal state, stress and material parameters are ground-truth diagnostics, not fitting supervision. Density, volume, initial geometry, boundaries and gravity are known simulation settings. All five trajectories pass the stored numerical checks. This is evidence of numerical consistency, not a guarantee of real-world material accuracy or of inverse identifiability.

The accompanying repository provides datasets.release download, which resolves an immutable commit and extracts the archives into stage2/homogeneous/<material>. manifest.json records archive sizes and counts. stage1/corpus.json is empty deliberately. The 32-GPU study must use the separately generated full v1 corpus, not this Stage-2-only snapshot.

Full v1 is now being generated and uploaded in bounded shards. Live progress is available in v1/progress.json; completed parts appear under v1/. The root manifest remains this historical snapshot until all 800 Stage-1 and 42 Stage-2 trajectories have been uploaded. Partial parts are not a complete training corpus.

Full v1: 800 homogeneous Stage-1 trajectories (200 material parameter samples, four excitations each) and 42 Stage-2 trajectories, all with 1,000 particles and 2,000 steps. Its generation recipes and launchers are provided in the code handoff. Node-local generation and immediate upload avoid retaining the whole raw Stage-1 corpus on the shared filesystem.

Synthetic dataset license: CC BY 4.0. This license does not relicense the code or simulation dependencies. No trained models, credentials, private observations, or derived fitting caches are included in this dataset.

Downloads last month
11