Datasets:

lsnu
/

octo-small-simpler-cube-stack-success-rtx4500

Error code:   StreamingRowsError
Exception:    CastError
Message:      Couldn't cast
schema_version: int64
started_utc: string
command: list<item: string>
  child 0, item: string
arguments: struct<env_id: string, num_envs: int64, allow_unvalidated_width: bool, num_episodes: int64, episode_ (... 396 chars omitted)
  child 0, env_id: string
  child 1, num_envs: int64
  child 2, allow_unvalidated_width: bool
  child 3, num_episodes: int64
  child 4, episode_start: int64
  child 5, env_seed: int64
  child 6, policy_seed: int64
  child 7, env_seed_mode: string
  child 8, policy_rng_mode: string
  child 9, repeats: int64
  child 10, shader: string
  child 11, record_activations: bool
  child 12, record_object_poses: bool
  child 13, save_video: bool
  child 14, video_max_episodes: int64
  child 15, video_fps: int64
  child 16, video_quality: int64
  child 17, sim_backend: string
  child 18, output_dir: string
  child 19, run_name: string
  child 20, fail_on_repeat_disagreement: bool
  child 21, fail_on_score_disagreement: bool
batches: list<item: struct<batch_index: int64, episode_ids: list<item: int64>, env_seeds: list<item: int64>,  (... 19 chars omitted)
  child 0, item: struct<batch_index: int64, episode_ids: list<item: int64>, env_seeds: list<item: int64>, policy_seed (... 7 chars omitted)
      child 0, batch_index: int64
      child 1, episode_ids: list<item: int64>
          child 0, item: int64
      child 2, env_seeds: list<item: int64>
          child 0, item: int64
      child 3, policy_seed: null
determinism_scope: struct<episode_ids_frozen: bool, per_
...
d 2, scripts/_env.sh: string
hardware: struct<nvidia_smi: string, jax_devices: list<item: string>>
  child 0, nvidia_smi: string
  child 1, jax_devices: list<item: string>
      child 0, item: string
determinism_environment: struct<CUDA_VISIBLE_DEVICES: null, CUBLAS_WORKSPACE_CONFIG: null, PYTHONHASHSEED: string, TF_DETERMI (... 73 chars omitted)
  child 0, CUDA_VISIBLE_DEVICES: null
  child 1, CUBLAS_WORKSPACE_CONFIG: null
  child 2, PYTHONHASHSEED: string
  child 3, TF_DETERMINISTIC_OPS: null
  child 4, XLA_FLAGS: null
  child 5, XLA_PYTHON_CLIENT_PREALLOCATE: string
completed_utc: string
artifact_sha256: struct<episodes.jsonl: string, summary.json: string>
  child 0, episodes.jsonl: string
  child 1, summary.json: string
metrics: struct<consecutive_grasp: bool, elapsed_steps: int64, is_src_obj_grasped: bool, moved_correct_obj: b (... 63 chars omitted)
  child 0, consecutive_grasp: bool
  child 1, elapsed_steps: int64
  child 2, is_src_obj_grasped: bool
  child 3, moved_correct_obj: bool
  child 4, moved_wrong_obj: bool
  child 5, src_on_target: bool
  child 6, success: bool
environment_seed: int64
activation_trace_sha256: null
policy_rng_before_sha256: string
episode_id: int64
batch_slot: int64
initial_state_sha256: string
object_pose_trace_sha256: string
elapsed_steps: int64
env_id: string
action_trace_sha256: string
batch_index: int64
final_state_sha256: string
instruction: string
video_path: string
repeat: int64
policy_batch_seed: null
global_camera_trace_sha256: string
to
{'action_trace_sha256': Value('string'), 'activation_trace_sha256': Value('null'), 'batch_index': Value('int64'), 'batch_slot': Value('int64'), 'elapsed_steps': Value('int64'), 'env_id': Value('string'), 'environment_seed': Value('int64'), 'episode_id': Value('int64'), 'final_state_sha256': Value('string'), 'global_camera_trace_sha256': Value('string'), 'initial_state_sha256': Value('string'), 'instruction': Value('string'), 'metrics': {'consecutive_grasp': Value('bool'), 'elapsed_steps': Value('int64'), 'is_src_obj_grasped': Value('bool'), 'moved_correct_obj': Value('bool'), 'moved_wrong_obj': Value('bool'), 'src_on_target': Value('bool'), 'success': Value('bool')}, 'object_pose_trace_sha256': Value('string'), 'policy_batch_seed': Value('null'), 'policy_rng_before_sha256': Value('string'), 'repeat': Value('int64'), 'video_path': Value('string')}
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 489, 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 2818, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2355, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2380, 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 2369, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2297, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              schema_version: int64
              started_utc: string
              command: list<item: string>
                child 0, item: string
              arguments: struct<env_id: string, num_envs: int64, allow_unvalidated_width: bool, num_episodes: int64, episode_ (... 396 chars omitted)
                child 0, env_id: string
                child 1, num_envs: int64
                child 2, allow_unvalidated_width: bool
                child 3, num_episodes: int64
                child 4, episode_start: int64
                child 5, env_seed: int64
                child 6, policy_seed: int64
                child 7, env_seed_mode: string
                child 8, policy_rng_mode: string
                child 9, repeats: int64
                child 10, shader: string
                child 11, record_activations: bool
                child 12, record_object_poses: bool
                child 13, save_video: bool
                child 14, video_max_episodes: int64
                child 15, video_fps: int64
                child 16, video_quality: int64
                child 17, sim_backend: string
                child 18, output_dir: string
                child 19, run_name: string
                child 20, fail_on_repeat_disagreement: bool
                child 21, fail_on_score_disagreement: bool
              batches: list<item: struct<batch_index: int64, episode_ids: list<item: int64>, env_seeds: list<item: int64>,  (... 19 chars omitted)
                child 0, item: struct<batch_index: int64, episode_ids: list<item: int64>, env_seeds: list<item: int64>, policy_seed (... 7 chars omitted)
                    child 0, batch_index: int64
                    child 1, episode_ids: list<item: int64>
                        child 0, item: int64
                    child 2, env_seeds: list<item: int64>
                        child 0, item: int64
                    child 3, policy_seed: null
              determinism_scope: struct<episode_ids_frozen: bool, per_
              ...
              d 2, scripts/_env.sh: string
              hardware: struct<nvidia_smi: string, jax_devices: list<item: string>>
                child 0, nvidia_smi: string
                child 1, jax_devices: list<item: string>
                    child 0, item: string
              determinism_environment: struct<CUDA_VISIBLE_DEVICES: null, CUBLAS_WORKSPACE_CONFIG: null, PYTHONHASHSEED: string, TF_DETERMI (... 73 chars omitted)
                child 0, CUDA_VISIBLE_DEVICES: null
                child 1, CUBLAS_WORKSPACE_CONFIG: null
                child 2, PYTHONHASHSEED: string
                child 3, TF_DETERMINISTIC_OPS: null
                child 4, XLA_FLAGS: null
                child 5, XLA_PYTHON_CLIENT_PREALLOCATE: string
              completed_utc: string
              artifact_sha256: struct<episodes.jsonl: string, summary.json: string>
                child 0, episodes.jsonl: string
                child 1, summary.json: string
              metrics: struct<consecutive_grasp: bool, elapsed_steps: int64, is_src_obj_grasped: bool, moved_correct_obj: b (... 63 chars omitted)
                child 0, consecutive_grasp: bool
                child 1, elapsed_steps: int64
                child 2, is_src_obj_grasped: bool
                child 3, moved_correct_obj: bool
                child 4, moved_wrong_obj: bool
                child 5, src_on_target: bool
                child 6, success: bool
              environment_seed: int64
              activation_trace_sha256: null
              policy_rng_before_sha256: string
              episode_id: int64
              batch_slot: int64
              initial_state_sha256: string
              object_pose_trace_sha256: string
              elapsed_steps: int64
              env_id: string
              action_trace_sha256: string
              batch_index: int64
              final_state_sha256: string
              instruction: string
              video_path: string
              repeat: int64
              policy_batch_seed: null
              global_camera_trace_sha256: string
              to
              {'action_trace_sha256': Value('string'), 'activation_trace_sha256': Value('null'), 'batch_index': Value('int64'), 'batch_slot': Value('int64'), 'elapsed_steps': Value('int64'), 'env_id': Value('string'), 'environment_seed': Value('int64'), 'episode_id': Value('int64'), 'final_state_sha256': Value('string'), 'global_camera_trace_sha256': Value('string'), 'initial_state_sha256': Value('string'), 'instruction': Value('string'), 'metrics': {'consecutive_grasp': Value('bool'), 'elapsed_steps': Value('int64'), 'is_src_obj_grasped': Value('bool'), 'moved_correct_obj': Value('bool'), 'moved_wrong_obj': Value('bool'), 'src_on_target': Value('bool'), 'success': Value('bool')}, 'object_pose_trace_sha256': Value('string'), 'policy_batch_seed': Value('null'), 'policy_rng_before_sha256': Value('string'), 'repeat': Value('int64'), 'video_path': Value('string')}
              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.

Octo-Small SIMPLER cube-stack success

This is a compact, auditable ManiSkill 3 evaluation artifact for rail-berkeley/octo-small on SIMPLER's BridgeData-v2 digital twin task:

Environment: StackGreenCubeOnYellowCubeBakedTexInScene-v1
Instruction: stack the green block on the yellow block
Robot/policy setup: WidowX Bridge / Octo-Small
Hardware: NVIDIA RTX A4500, 20 GB
Date: 2026-07-22 UTC

Videos

Episode	Terminal result	Notes
48	Failure	Sustained grasp, but the green cube was not on the yellow cube at timeout.
49	Success	Green cube grasped, placed on the yellow cube, and released into a stable stack.

The primary artifact is videos/success/episode_000049.mp4. Episode 48 is included as the immediately preceding failure in the deterministic policy stream.

Both videos are H.264, 640x480, 10 fps, 61 frames, and 6.1 seconds long. Visual inspection confirms that episode 49 ends with an open gripper and the green cube resting on the yellow cube. The pose trace independently shows that the green cube moved by at most 0.0191 mm over the final ten control steps.

Evaluation protocol

The evaluator ran one 48-environment GPU batch covering episode IDs 48-95, then reconstructed the same batch for an exact replay audit. The batch produced three terminal successes (episodes 49, 53, and 77); video encoding was intentionally limited to episodes 48 and 49, stopping at the first success.

Environment seeds: environment_seed = episode_id
Octo seed: 0
Octo RNG mode: publication-compatible continuous stream, advanced past the preceding episode-0-47 batch
Episode horizon: 60 control steps
Parallel environments: 48
Repeats: 2
Batch success: 3/48 (6.25%)
Exact replay: 48/48 initial states, action traces, final states, terminal metrics, and object-pose traces matched; 2/2 encoded camera traces matched

Reproduction command:

scripts/eval_trusted \
  -e StackGreenCubeOnYellowCubeBakedTexInScene-v1 \
  --num-envs 48 \
  --num-episodes 48 \
  --episode-start 48 \
  --env-seed 0 \
  --policy-seed 0 \
  --repeats 2 \
  --save-video \
  --video-max-episodes 2 \
  --record-object-poses \
  --run-name cube-stack-until-success-e48-n48

Provenance

provenance/ contains the complete 48-episode records for both repeats, the run manifest, summary, and the first repeat's per-step robot/object pose archive. The manifest records package versions, repository commits and diffs, model revision, hardware, seeds, command, and evaluator hashes. SHA256SUMS covers every uploaded artifact.

The success label is ManiSkill's terminal src_on_target predicate for this task, which requires the green cube to be geometrically above and in contact with the yellow cube. This artifact is intended for evaluation analysis rather than policy training.

Downloads last month: 74

Total file size:

589 kB