Dataset Viewer
Split (1)
train
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
criterion: string
candidates: list<item: struct<checkpoint: string, report: string, epoch: int64, mean_nll: double, coarse_nll: do (... 390 chars omitted)
  child 0, item: struct<checkpoint: string, report: string, epoch: int64, mean_nll: double, coarse_nll: double, fine_ (... 378 chars omitted)
      child 0, checkpoint: string
      child 1, report: string
      child 2, epoch: int64
      child 3, mean_nll: double
      child 4, coarse_nll: double
      child 5, fine_nll: double
      child 6, shuffled_mean_nll: double
      child 7, shuffled_minus_own: double
      child 8, by_source: struct<anet: struct<segments_per_scale: int64, coarse_nll: double, fine_nll: double, mean_nll: doubl (... 198 chars omitted)
          child 0, anet: struct<segments_per_scale: int64, coarse_nll: double, fine_nll: double, mean_nll: double>
              child 0, segments_per_scale: int64
              child 1, coarse_nll: double
              child 2, fine_nll: double
              child 3, mean_nll: double
          child 1, cha: struct<segments_per_scale: int64, coarse_nll: double, fine_nll: double, mean_nll: double>
              child 0, segments_per_scale: int64
              child 1, coarse_nll: double
              child 2, fine_nll: double
              child 3, mean_nll: double
          child 2, didemo: struct<segments_per_scale: int64, coarse_nll: double, fine_nll: double, mean_nll: double>
              child 0, segments_per_scale: int64
              child 1, coarse_nll: double

...
mp_steps: int64
  child 56, pointer_loss_weight: double
  child 57, answer_loss_weight: double
started_at: double
data: struct<selected: int64, train: int64, val: int64, missing: struct<>, excluded_rows: struct<>, exclud (... 183 chars omitted)
  child 0, selected: int64
  child 1, train: int64
  child 2, val: int64
  child 3, missing: struct<>
  child 4, excluded_rows: struct<>
  child 5, excluded_case: struct<>
  child 6, excluded_videos: struct<llava: int64, lvdb: int64>
      child 0, llava: int64
      child 1, lvdb: int64
  child 7, bad_answer: int64
  child 8, train_case: struct<C: int64, B: int64, A: int64>
      child 0, C: int64
      child 1, B: int64
      child 2, A: int64
  child 9, train_format: struct<mc: int64, oe: int64>
      child 0, mc: int64
      child 1, oe: int64
writer_checkpoint: string
writer_sha256: string
optimizer_groups: list<item: struct<name: string, lr: double, parameters: int64>>
  child 0, item: struct<name: string, lr: double, parameters: int64>
      child 0, name: string
      child 1, lr: double
      child 2, parameters: int64
batch_contract: struct<world_size: int64, microbatch_per_gpu: int64, gradient_accumulation: int64, effective_global_ (... 69 chars omitted)
  child 0, world_size: int64
  child 1, microbatch_per_gpu: int64
  child 2, gradient_accumulation: int64
  child 3, effective_global_batch: int64
  child 4, epoch_samples: int64
  child 5, optimizer_steps_per_epoch: int64
val_qids: list<item: string>
  child 0, item: string
to
{'format': Value('string'), 'action_protocol': Value('string'), 'design_version': Value('string'), 'position_contract': Value('string'), 'lora_target_modules': List(Value('string')), 'epoch_data_policy': {'format': Value('string'), 'source_weights_applied': Value('bool'), 'source_rows': Value('int64'), 'epoch_rows': Value('int64'), 'dropped_global_batch_remainder': Value('int64')}, 'started_at': Value('float64'), 'writer_checkpoint': Value('string'), 'writer_sha256': Value('string'), 'args': {'model': Value('string'), 'traces': Value('string'), 'capfilter': Value('string'), 'writer_checkpoint': Value('string'), 'preflight_manifest': Value('string'), 'resume': Value('string'), 'llava_kv_root': Value('string'), 'lvdb_kv_root': Value('string'), 'video_exclusions': Value('string'), 'raw_oracle_policy': Value('string'), 'raw_top_k': Value('int64'), 'source_weights': Value('string'), 'val_percent': Value('int64'), 'allow_partial_cache': Value('bool'), 'limit_train': Value('int64'), 'epochs': Value('int64'), 'global_batch': Value('int64'), 'expected_world_size': Value('int64'), 'lr': Value('float64'), 'lora_lr': Value('float64'), 'lora_r': Value('int64'), 'lora_alpha': Value('int64'), 'max_grad_norm': Value('float64'), 'seed': Value('int64'), 'log_every': Value('int64'), 'memory_position_mode': Value('string'), 'case_a_memory_mixture': Value('string'), 'case_a_hybrid_fine_count': Value('int64'), 'replay_probe': Value('bool'), 'replay_probe_only': Value('bool'), 'oe_answer_max_tokens
...
18_reader_periphery.py': Value('string'), 'me18_reader_raw_budget.py': Value('string'), 'me18_reader_replay.py': Value('string'), 'me18_reader_sft_core.py': Value('string'), 'me18_reader_trajectory.py': Value('string'), 'me18_source_provenance.py': Value('string'), 'me18_video_exclusions.py': Value('string')}}, 'video_exclusions_provenance': Value('null'), 'trainable_parameters': {'backbone_trainable_kind': Value('string'), 'lora_tensors': Value('int64'), 'lora_parameters': Value('int64'), 'protocol_row_tensors': Value('int64'), 'protocol_row_parameters': Value('int64'), 'periphery_tensors': Value('int64'), 'periphery_parameters': Value('int64'), 'mixer_shape': List(Value('int64')), 'mixer_scalars': Value('int64')}, 'optimizer_groups': List({'name': Value('string'), 'lr': Value('float64'), 'parameters': Value('int64')}), 'batch_contract': {'world_size': Value('int64'), 'microbatch_per_gpu': Value('int64'), 'gradient_accumulation': Value('int64'), 'effective_global_batch': Value('int64'), 'epoch_samples': Value('int64'), 'optimizer_steps_per_epoch': Value('int64')}, 'data': {'selected': Value('int64'), 'train': Value('int64'), 'val': Value('int64'), 'missing': {}, 'excluded_rows': {}, 'excluded_case': {}, 'excluded_videos': {'llava': Value('int64'), 'lvdb': Value('int64')}, 'bad_answer': Value('int64'), 'train_case': {'C': Value('int64'), 'B': Value('int64'), 'A': Value('int64')}, 'train_format': {'mc': Value('int64'), 'oe': Value('int64')}}, 'val_qids': List(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
              criterion: string
              candidates: list<item: struct<checkpoint: string, report: string, epoch: int64, mean_nll: double, coarse_nll: do (... 390 chars omitted)
                child 0, item: struct<checkpoint: string, report: string, epoch: int64, mean_nll: double, coarse_nll: double, fine_ (... 378 chars omitted)
                    child 0, checkpoint: string
                    child 1, report: string
                    child 2, epoch: int64
                    child 3, mean_nll: double
                    child 4, coarse_nll: double
                    child 5, fine_nll: double
                    child 6, shuffled_mean_nll: double
                    child 7, shuffled_minus_own: double
                    child 8, by_source: struct<anet: struct<segments_per_scale: int64, coarse_nll: double, fine_nll: double, mean_nll: doubl (... 198 chars omitted)
                        child 0, anet: struct<segments_per_scale: int64, coarse_nll: double, fine_nll: double, mean_nll: double>
                            child 0, segments_per_scale: int64
                            child 1, coarse_nll: double
                            child 2, fine_nll: double
                            child 3, mean_nll: double
                        child 1, cha: struct<segments_per_scale: int64, coarse_nll: double, fine_nll: double, mean_nll: double>
                            child 0, segments_per_scale: int64
                            child 1, coarse_nll: double
                            child 2, fine_nll: double
                            child 3, mean_nll: double
                        child 2, didemo: struct<segments_per_scale: int64, coarse_nll: double, fine_nll: double, mean_nll: double>
                            child 0, segments_per_scale: int64
                            child 1, coarse_nll: double
              
              ...
              mp_steps: int64
                child 56, pointer_loss_weight: double
                child 57, answer_loss_weight: double
              started_at: double
              data: struct<selected: int64, train: int64, val: int64, missing: struct<>, excluded_rows: struct<>, exclud (... 183 chars omitted)
                child 0, selected: int64
                child 1, train: int64
                child 2, val: int64
                child 3, missing: struct<>
                child 4, excluded_rows: struct<>
                child 5, excluded_case: struct<>
                child 6, excluded_videos: struct<llava: int64, lvdb: int64>
                    child 0, llava: int64
                    child 1, lvdb: int64
                child 7, bad_answer: int64
                child 8, train_case: struct<C: int64, B: int64, A: int64>
                    child 0, C: int64
                    child 1, B: int64
                    child 2, A: int64
                child 9, train_format: struct<mc: int64, oe: int64>
                    child 0, mc: int64
                    child 1, oe: int64
              writer_checkpoint: string
              writer_sha256: string
              optimizer_groups: list<item: struct<name: string, lr: double, parameters: int64>>
                child 0, item: struct<name: string, lr: double, parameters: int64>
                    child 0, name: string
                    child 1, lr: double
                    child 2, parameters: int64
              batch_contract: struct<world_size: int64, microbatch_per_gpu: int64, gradient_accumulation: int64, effective_global_ (... 69 chars omitted)
                child 0, world_size: int64
                child 1, microbatch_per_gpu: int64
                child 2, gradient_accumulation: int64
                child 3, effective_global_batch: int64
                child 4, epoch_samples: int64
                child 5, optimizer_steps_per_epoch: int64
              val_qids: list<item: string>
                child 0, item: string
              to
              {'format': Value('string'), 'action_protocol': Value('string'), 'design_version': Value('string'), 'position_contract': Value('string'), 'lora_target_modules': List(Value('string')), 'epoch_data_policy': {'format': Value('string'), 'source_weights_applied': Value('bool'), 'source_rows': Value('int64'), 'epoch_rows': Value('int64'), 'dropped_global_batch_remainder': Value('int64')}, 'started_at': Value('float64'), 'writer_checkpoint': Value('string'), 'writer_sha256': Value('string'), 'args': {'model': Value('string'), 'traces': Value('string'), 'capfilter': Value('string'), 'writer_checkpoint': Value('string'), 'preflight_manifest': Value('string'), 'resume': Value('string'), 'llava_kv_root': Value('string'), 'lvdb_kv_root': Value('string'), 'video_exclusions': Value('string'), 'raw_oracle_policy': Value('string'), 'raw_top_k': Value('int64'), 'source_weights': Value('string'), 'val_percent': Value('int64'), 'allow_partial_cache': Value('bool'), 'limit_train': Value('int64'), 'epochs': Value('int64'), 'global_batch': Value('int64'), 'expected_world_size': Value('int64'), 'lr': Value('float64'), 'lora_lr': Value('float64'), 'lora_r': Value('int64'), 'lora_alpha': Value('int64'), 'max_grad_norm': Value('float64'), 'seed': Value('int64'), 'log_every': Value('int64'), 'memory_position_mode': Value('string'), 'case_a_memory_mixture': Value('string'), 'case_a_hybrid_fine_count': Value('int64'), 'replay_probe': Value('bool'), 'replay_probe_only': Value('bool'), 'oe_answer_max_tokens
              ...
              18_reader_periphery.py': Value('string'), 'me18_reader_raw_budget.py': Value('string'), 'me18_reader_replay.py': Value('string'), 'me18_reader_sft_core.py': Value('string'), 'me18_reader_trajectory.py': Value('string'), 'me18_source_provenance.py': Value('string'), 'me18_video_exclusions.py': Value('string')}}, 'video_exclusions_provenance': Value('null'), 'trainable_parameters': {'backbone_trainable_kind': Value('string'), 'lora_tensors': Value('int64'), 'lora_parameters': Value('int64'), 'protocol_row_tensors': Value('int64'), 'protocol_row_parameters': Value('int64'), 'periphery_tensors': Value('int64'), 'periphery_parameters': Value('int64'), 'mixer_shape': List(Value('int64')), 'mixer_scalars': Value('int64')}, 'optimizer_groups': List({'name': Value('string'), 'lr': Value('float64'), 'parameters': Value('int64')}), 'batch_contract': {'world_size': Value('int64'), 'microbatch_per_gpu': Value('int64'), 'gradient_accumulation': Value('int64'), 'effective_global_batch': Value('int64'), 'epoch_samples': Value('int64'), 'optimizer_steps_per_epoch': Value('int64')}, 'data': {'selected': Value('int64'), 'train': Value('int64'), 'val': Value('int64'), 'missing': {}, 'excluded_rows': {}, 'excluded_case': {}, 'excluded_videos': {'llava': Value('int64'), 'lvdb': Value('int64')}, 'bad_answer': Value('int64'), 'train_case': {'C': Value('int64'), 'B': Value('int64'), 'A': Value('int64')}, 'train_format': {'mc': Value('int64'), 'oe': Value('int64')}}, 'val_qids': List(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.
YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
MemLove artifact bundle

This dataset repository contains the selected MemLove Writer/Reader checkpoints and the cached gist tensors used by the current experiments.
Layout

artifacts/checkpoints/writer/: selected Writer checkpoint and selection metadata.
artifacts/checkpoints/reader/: primary relocate-K + sequential-Q SFT epoch-1 Reader checkpoint and run metadata.
artifacts/gists/native_v2/: per-video me18-native-kv-gist-cache-v2 files.
artifacts/benchmarks/<benchmark>/joint_parts/: benchmark me18-joint-masked-kv-gist-cache-v1 files.
The bundle intentionally excludes old-format native-v1 gist caches, optimizer states, raw videos, source dataset archives, training/evaluation logs, and unrelated experiment checkpoints.
Native-v2 postflight found 30,211 of 30,215 locally available videos cached. The four videos that still fail after retry were approved as exclusions. They are documented in MANIFEST.md and in the uploaded postflight report/exclusion manifests.
See MANIFEST.md for source paths, file counts, sizes, and checkpoint hashes.
Downloads last month: -
Total file size:
226 GB