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
tokenizer: string
source: string
subsets: struct<m4_hourly: struct<n_series: int64, n_tokens: int64, leaking: bool>>
  child 0, m4_hourly: struct<n_series: int64, n_tokens: int64, leaking: bool>
      child 0, n_series: int64
      child 1, n_tokens: int64
      child 2, leaking: bool
n_series: int64
shards: int64
n_tokens: int64
subset: string
leaking: bool
to
{'tokenizer': Value('string'), 'source': Value('string'), 'subset': Value('string'), 'n_series': Value('int64'), 'n_tokens': Value('int64'), 'shards': Value('int64'), 'leaking': 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 478, 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
              tokenizer: string
              source: string
              subsets: struct<m4_hourly: struct<n_series: int64, n_tokens: int64, leaking: bool>>
                child 0, m4_hourly: struct<n_series: int64, n_tokens: int64, leaking: bool>
                    child 0, n_series: int64
                    child 1, n_tokens: int64
                    child 2, leaking: bool
              n_series: int64
              shards: int64
              n_tokens: int64
              subset: string
              leaking: bool
              to
              {'tokenizer': Value('string'), 'source': Value('string'), 'subset': Value('string'), 'n_series': Value('int64'), 'n_tokens': Value('int64'), 'shards': Value('int64'), 'leaking': 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.

Resonance-TS — E8-Tokenized Time-Series Corpus

A large collection of public time-series datasets, pre-tokenized with a collapse-proof E8 lattice tokenizer, so you can train time-series foundation models straight from compact tokens — without re-downloading terabytes of raw data or building a tokenizer yourself.

Made public so it's useful to the community. If it saves you the download + tokenize step, that's the point.

What's inside

Each source dataset is tokenized into compact uint16 token packs. The tokens are near-lossless and collapse-proof (unlike the uniform value-binning used by some models, which wastes/degenerates codes).

The tokenizer — e8rvq_p16_s4_v2

series --instance-norm--> clip[-3,3] --patch(16)--> 2 blocks of 8-D --4-stage residual E8 quantization-->
        (num_patches, 2 blocks, 4 stages) uint16 token ids
  • E8 lattice = densest packing / rate-distortion optimal in 8-D → no codebook collapse.
  • Residual (RVQ) = 4 refinement stages → clean-signal reconstruction NMSE ≈ 0.5% (near-lossless).
  • Training-free & deterministic → identical tokens across every corpus, zero setup, reproducible.
  • 0.5 tokens/point (2× compression) · vocab 26,641.
  • Tokenizer + detokenize() code: github.com/QLNI/resonance-ts (tokenizer_e8.py).

Layout

e8rvq_p16_s4_v2/<source>__<dataset>/
    <subset>.npz        # object array of (num_patches, 2, 4) uint16 token matrices, one per series
    manifest.json       # per-subset: n_series, n_tokens, and a `leaking` flag

Sources

Public time-series archives (each retains its original license — cite the source dataset when you use it): autogluon/chronos_datasets, Salesforce/GiftEvalPretrain, Salesforce/lotsa_data, Monash, UCR/UEA, FRED, UCI, plus ~15% synthetic (KernelSynth-style).

Leakage flags (important for GIFT-Eval)

Any dataset overlapping the GIFT-Eval test set is tagged "leaking": true in its manifest.json. If you evaluate on GIFT-Eval, exclude the leaking subsets from training — otherwise your score is contaminated. The GiftEvalPretrain-derived packs are the non-leaking core.

Usage

import numpy as np
from huggingface_hub import hf_hub_download

f = hf_hub_download("NODEMIND/resonance-ts-tokens",
                    "e8rvq_p16_s4_v2/Monash-University__monash_tsf/traffic.npz",
                    repo_type="dataset")
pack = np.load(f, allow_pickle=True)
tokens = pack["arr_0"]                      # (num_patches, 2, 4) uint16  — first series
# reconstruct values:  from tokenizer_e8 import E8Tokenizer;  E8Tokenizer().detokenize(tokens)

Citation / attribution

Built for the Resonance-TS foundation-model project by NODEMIND (Sai Kiran Bathula). Please also cite the underlying source datasets. Tokenizer method: E8 residual-lattice quantization.

Downloads last month
139