Datasets:

AETHORIA-AI
/

TR-HASH-Pretraining-125B-Agentic-32K

Tasks:

Text Generation

Languages:

English

French

License:

Dataset card Files Files and versions

xet

Community

Dataset Viewer

Duplicate

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
schema: string
config_sha256: string
tokenizer_sha256: string
requested_tokens: int64
actual_tokens: int64
seq_len: int64
materialization_mode: string
exact_document_deduplication: bool
benchmark_decontamination: bool
agentic_signal_filtering: bool
protected_prompt_count: int64
protected_index_sha256: string
sources: list<item: struct<name: string, bucket: string, selection: string, weight: double, requested_source_ (... 441 chars omitted)
  child 0, item: struct<name: string, bucket: string, selection: string, weight: double, requested_source_tokens: int (... 429 chars omitted)
      child 0, name: string
      child 1, bucket: string
      child 2, selection: string
      child 3, weight: double
      child 4, requested_source_tokens: int64
      child 5, license_audit: string
      child 6, input: struct<source_type: string, dataset_id: string, config_name: string, split: string, revision: string (... 1 chars omitted)
          child 0, source_type: string
          child 1, dataset_id: string
          child 2, config_name: string
          child 3, split: string
          child 4, revision: string
      child 7, target_rows: int64
      child 8, rows_done: int64
      child 9, scanned: int64
      child 10, source_tokens: int64
      child 11, counters: struct<retained: int64, scanned: int64>
          child 0, retained: int64
          child 1, scanned: int64
      child 12, signal_counts: struct<>
      child 13, shards: list<item: struct<source: string, shard_index: i
...
4, (... 31 chars omitted)
          child 0, item: struct<source: string, shard_index: int64, repo_path: string, rows: int64, tokens: int64, bytes: int (... 19 chars omitted)
              child 0, source: string
              child 1, shard_index: int64
              child 2, repo_path: string
              child 3, rows: int64
              child 4, tokens: int64
              child 5, bytes: int64
              child 6, sha256: string
invariants: struct<replay: bool, each_packed_row_consumed_once: bool, preserve_source_proportions_within_bucket: (... 69 chars omitted)
  child 0, replay: bool
  child 1, each_packed_row_consumed_once: bool
  child 2, preserve_source_proportions_within_bucket: bool
  child 3, require_manifest_sha256: bool
  child 4, require_tokenizer_sha256: bool
version: int64
total_tokens: int64
phases: list<item: struct<name: string, order: int64, target_tokens: int64, bucket_tokens: struct<foundation (... 86 chars omitted)
  child 0, item: struct<name: string, order: int64, target_tokens: int64, bucket_tokens: struct<foundation: int64, ag (... 74 chars omitted)
      child 0, name: string
      child 1, order: int64
      child 2, target_tokens: int64
      child 3, bucket_tokens: struct<foundation: int64, agentic: int64>
          child 0, foundation: int64
          child 1, agentic: int64
      child 4, bucket_shares: struct<foundation: double, agentic: double>
          child 0, foundation: double
          child 1, agentic: double
description: string
to
{'version': Value('int64'), 'schema': Value('string'), 'description': Value('string'), 'total_tokens': Value('int64'), 'phases': List({'name': Value('string'), 'order': Value('int64'), 'target_tokens': Value('int64'), 'bucket_tokens': {'foundation': Value('int64'), 'agentic': Value('int64')}, 'bucket_shares': {'foundation': Value('float64'), 'agentic': Value('float64')}}), 'invariants': {'replay': Value('bool'), 'each_packed_row_consumed_once': Value('bool'), 'preserve_source_proportions_within_bucket': Value('bool'), 'require_manifest_sha256': Value('bool'), 'require_tokenizer_sha256': Value('bool')}}
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
              schema: string
              config_sha256: string
              tokenizer_sha256: string
              requested_tokens: int64
              actual_tokens: int64
              seq_len: int64
              materialization_mode: string
              exact_document_deduplication: bool
              benchmark_decontamination: bool
              agentic_signal_filtering: bool
              protected_prompt_count: int64
              protected_index_sha256: string
              sources: list<item: struct<name: string, bucket: string, selection: string, weight: double, requested_source_ (... 441 chars omitted)
                child 0, item: struct<name: string, bucket: string, selection: string, weight: double, requested_source_tokens: int (... 429 chars omitted)
                    child 0, name: string
                    child 1, bucket: string
                    child 2, selection: string
                    child 3, weight: double
                    child 4, requested_source_tokens: int64
                    child 5, license_audit: string
                    child 6, input: struct<source_type: string, dataset_id: string, config_name: string, split: string, revision: string (... 1 chars omitted)
                        child 0, source_type: string
                        child 1, dataset_id: string
                        child 2, config_name: string
                        child 3, split: string
                        child 4, revision: string
                    child 7, target_rows: int64
                    child 8, rows_done: int64
                    child 9, scanned: int64
                    child 10, source_tokens: int64
                    child 11, counters: struct<retained: int64, scanned: int64>
                        child 0, retained: int64
                        child 1, scanned: int64
                    child 12, signal_counts: struct<>
                    child 13, shards: list<item: struct<source: string, shard_index: i
              ...
              4, (... 31 chars omitted)
                        child 0, item: struct<source: string, shard_index: int64, repo_path: string, rows: int64, tokens: int64, bytes: int (... 19 chars omitted)
                            child 0, source: string
                            child 1, shard_index: int64
                            child 2, repo_path: string
                            child 3, rows: int64
                            child 4, tokens: int64
                            child 5, bytes: int64
                            child 6, sha256: string
              invariants: struct<replay: bool, each_packed_row_consumed_once: bool, preserve_source_proportions_within_bucket: (... 69 chars omitted)
                child 0, replay: bool
                child 1, each_packed_row_consumed_once: bool
                child 2, preserve_source_proportions_within_bucket: bool
                child 3, require_manifest_sha256: bool
                child 4, require_tokenizer_sha256: bool
              version: int64
              total_tokens: int64
              phases: list<item: struct<name: string, order: int64, target_tokens: int64, bucket_tokens: struct<foundation (... 86 chars omitted)
                child 0, item: struct<name: string, order: int64, target_tokens: int64, bucket_tokens: struct<foundation: int64, ag (... 74 chars omitted)
                    child 0, name: string
                    child 1, order: int64
                    child 2, target_tokens: int64
                    child 3, bucket_tokens: struct<foundation: int64, agentic: int64>
                        child 0, foundation: int64
                        child 1, agentic: int64
                    child 4, bucket_shares: struct<foundation: double, agentic: double>
                        child 0, foundation: double
                        child 1, agentic: double
              description: string
              to
              {'version': Value('int64'), 'schema': Value('string'), 'description': Value('string'), 'total_tokens': Value('int64'), 'phases': List({'name': Value('string'), 'order': Value('int64'), 'target_tokens': Value('int64'), 'bucket_tokens': {'foundation': Value('int64'), 'agentic': Value('int64')}, 'bucket_shares': {'foundation': Value('float64'), 'agentic': Value('float64')}}), 'invariants': {'replay': Value('bool'), 'each_packed_row_consumed_once': Value('bool'), 'preserve_source_proportions_within_bucket': Value('bool'), 'require_manifest_sha256': Value('bool'), 'require_tokenizer_sha256': Value('bool')}}
              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.

TR-HASH Pretraining 125B — Agentic 32K

Private, source-curated pretraining artifact for the TR-HASH Agentic 32K model line. It contains 125B packed token exposures: 75B foundation and 50B agentic/procedural content.

The corpus uses the immutable, validated 32,000-ID revision of AETHORIA-AI/TR-HASH-Tokenizer-32K-Agentic. It is not compatible with the older TR-HASH 32K tokenizer.

Composition

Bucket	Tokens	Purpose
Foundation	75B	English and French knowledge, educational web text, math and synthetic textbooks
Agentic	50B	Educational code, procedures, debugging, verification, planning, math reasoning and capped tool-use trajectories

All source repositories, configurations, revisions, token budgets, and license audit notes are pinned in the published _metadata/config.json.

Materialization disclosure

This is the high-throughput source-curated direct build. Documents are read from curated upstream subsets and tokenized directly. The build intentionally does not claim per-document quality filtering, agentic-signal filtering, benchmark decontamination, or global exact-document deduplication.

Each one-billion-token uint16 shard is uploaded, checked against its remote size and SHA-256, committed to restart state, and only then deleted locally. The dataset remains private until source licenses, shard hashes, token budgets, and this materialization disclosure have been audited.

Curriculum

The runtime plan consumes every packed row once:

foundation-first: 60B foundation + 15B agentic.
agentic-intensification: 15B foundation + 35B agentic.

Shard boundaries are aligned to the phase split. unique_tokens and trained_tokens describe packed token positions; source documents may repeat because global deduplication is disabled in this direct build.

Downloads last month: 543

Total file size:

250 GB

Models trained or fine-tuned on AETHORIA-AI/TR-HASH-Pretraining-125B-Agentic-32K

AETHORIA-AI/TR-HASH-MoE-100M-125B-Agentic-Pretraining

Text Generation • Updated 1 day ago • 3.18k

AETHORIA-AI/TR-HASH-MoE-100M-70B-Agentic-Refinement

Text Generation • Updated about 13 hours ago • 1.34k