Datasets:

EchoLabs33
/

regulated-asset-tensor-hxq

Error code:   StreamingRowsError
Exception:    CastError
Message:      Couldn't cast
schema_version: string
asset_type: string
description: string
features: list<item: struct<index: int64, name: string, type: string, distribution: string, range: string>>
  child 0, item: struct<index: int64, name: string, type: string, distribution: string, range: string>
      child 0, index: int64
      child 1, name: string
      child 2, type: string
      child 3, distribution: string
      child 4, range: string
codec: string
provenance: struct<synthetic_only: bool, no_real_pii: bool, no_real_transactions: bool, policy_version: string,  (... 12 chars omitted)
  child 0, synthetic_only: bool
  child 1, no_real_pii: bool
  child 2, no_real_transactions: bool
  child 3, policy_version: string
  child 4, seed: int64
title: string
n_events: int64
asset_source: string
seed: int64
element_count: int64
dtype: string
synthetic_only: bool
cost: struct<wall_time_s: double, cpu_time_s: double, peak_memory_mb: double, python_version: string, nump (... 84 chars omitted)
  child 0, wall_time_s: double
  child 1, cpu_time_s: double
  child 2, peak_memory_mb: double
  child 3, python_version: string
  child 4, numpy_version: string
  child 5, hostname: string
  child 6, timestamp_start: string
  child 7, timestamp_end: string
n_features: int64
fidelity: struct<cosine_similarity: double, cosine_min_per_feature: double, rms_error: double, max_abs_error:  (... 58 chars omitted)
  child 0, cosine_similarity: double
  child 1, cosine_min_per_feature: double
  child 2, rms_error: double
  child 3, max_abs_error: double
  child 4, mse: double
  child 5, gate_threshold: double
  child 6, gate: string
size: struct<original_bytes: int64, compressed_bytes: int64, compression_ratio: double>
  child 0, original_bytes: int64
  child 1, compressed_bytes: int64
  child 2, compression_ratio: double
no_real_transactions: bool
hashes: struct<sha256_original: string, sha256_decoded: string>
  child 0, sha256_original: string
  child 1, sha256_decoded: string
tensor_shape: list<item: int64>
  child 0, item: int64
receipt_id: string
bits_per_weight: int64
n_levels: int64
feature_names: list<item: string>
  child 0, item: string
policy_context: struct<policy_version: string, adapter_source: string, note: string>
  child 0, policy_version: string
  child 1, adapter_source: string
  child 2, note: string
per_feature_cosine: list<item: struct<feature: string, cosine: double>>
  child 0, item: struct<feature: string, cosine: double>
      child 0, feature: string
      child 1, cosine: double
group_size: int64
no_real_pii: bool
to
{'receipt_id': Value('string'), 'title': Value('string'), 'asset_type': Value('string'), 'asset_source': Value('string'), 'synthetic_only': Value('bool'), 'no_real_pii': Value('bool'), 'no_real_transactions': Value('bool'), 'codec': Value('string'), 'group_size': Value('int64'), 'n_levels': Value('int64'), 'bits_per_weight': Value('int64'), 'tensor_shape': List(Value('int64')), 'n_events': Value('int64'), 'n_features': Value('int64'), 'feature_names': List(Value('string')), 'element_count': Value('int64'), 'dtype': Value('string'), 'seed': Value('int64'), 'fidelity': {'cosine_similarity': Value('float64'), 'cosine_min_per_feature': Value('float64'), 'rms_error': Value('float64'), 'max_abs_error': Value('float64'), 'mse': Value('float64'), 'gate_threshold': Value('float64'), 'gate': Value('string')}, 'per_feature_cosine': List({'feature': Value('string'), 'cosine': Value('float64')}), 'size': {'original_bytes': Value('int64'), 'compressed_bytes': Value('int64'), 'compression_ratio': Value('float64')}, 'hashes': {'sha256_original': Value('string'), 'sha256_decoded': Value('string')}, 'policy_context': {'policy_version': Value('string'), 'adapter_source': Value('string'), 'note': Value('string')}, 'schema_version': Value('string'), 'cost': {'wall_time_s': Value('float64'), 'cpu_time_s': Value('float64'), 'peak_memory_mb': Value('float64'), 'python_version': Value('string'), 'numpy_version': Value('string'), 'hostname': Value('string'), 'timestamp_start': Value('string'), 'timestamp_end': Value('string')}}
because column names don't match
Traceback:    Traceback (most recent call last):
                File "/src/services/worker/src/worker/utils.py", line 99, in get_rows_or_raise
                  return get_rows(
                         ^^^^^^^^^
                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 77, in get_rows
                  rows_plus_one = list(itertools.islice(ds, rows_max_number + 1))
                                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 2690, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 2227, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 2251, in _iter_arrow
                  for key, pa_table in self.ex_iterable._iter_arrow():
                                       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 494, in _iter_arrow
                  for key, pa_table in iterator:
                                       ^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 384, in _iter_arrow
                  for key, pa_table in self.generate_tables_fn(**gen_kwags):
                                       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/packaged_modules/json/json.py", line 295, in _generate_tables
                  self._cast_table(pa_table, json_field_paths=json_field_paths),
                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/packaged_modules/json/json.py", line 128, in _cast_table
                  pa_table = table_cast(pa_table, self.info.features.arrow_schema)
                             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/table.py", line 2321, in table_cast
                  return cast_table_to_schema(table, schema)
                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/table.py", line 2249, in cast_table_to_schema
                  raise CastError(
              datasets.table.CastError: Couldn't cast
              schema_version: string
              asset_type: string
              description: string
              features: list<item: struct<index: int64, name: string, type: string, distribution: string, range: string>>
                child 0, item: struct<index: int64, name: string, type: string, distribution: string, range: string>
                    child 0, index: int64
                    child 1, name: string
                    child 2, type: string
                    child 3, distribution: string
                    child 4, range: string
              codec: string
              provenance: struct<synthetic_only: bool, no_real_pii: bool, no_real_transactions: bool, policy_version: string,  (... 12 chars omitted)
                child 0, synthetic_only: bool
                child 1, no_real_pii: bool
                child 2, no_real_transactions: bool
                child 3, policy_version: string
                child 4, seed: int64
              title: string
              n_events: int64
              asset_source: string
              seed: int64
              element_count: int64
              dtype: string
              synthetic_only: bool
              cost: struct<wall_time_s: double, cpu_time_s: double, peak_memory_mb: double, python_version: string, nump (... 84 chars omitted)
                child 0, wall_time_s: double
                child 1, cpu_time_s: double
                child 2, peak_memory_mb: double
                child 3, python_version: string
                child 4, numpy_version: string
                child 5, hostname: string
                child 6, timestamp_start: string
                child 7, timestamp_end: string
              n_features: int64
              fidelity: struct<cosine_similarity: double, cosine_min_per_feature: double, rms_error: double, max_abs_error:  (... 58 chars omitted)
                child 0, cosine_similarity: double
                child 1, cosine_min_per_feature: double
                child 2, rms_error: double
                child 3, max_abs_error: double
                child 4, mse: double
                child 5, gate_threshold: double
                child 6, gate: string
              size: struct<original_bytes: int64, compressed_bytes: int64, compression_ratio: double>
                child 0, original_bytes: int64
                child 1, compressed_bytes: int64
                child 2, compression_ratio: double
              no_real_transactions: bool
              hashes: struct<sha256_original: string, sha256_decoded: string>
                child 0, sha256_original: string
                child 1, sha256_decoded: string
              tensor_shape: list<item: int64>
                child 0, item: int64
              receipt_id: string
              bits_per_weight: int64
              n_levels: int64
              feature_names: list<item: string>
                child 0, item: string
              policy_context: struct<policy_version: string, adapter_source: string, note: string>
                child 0, policy_version: string
                child 1, adapter_source: string
                child 2, note: string
              per_feature_cosine: list<item: struct<feature: string, cosine: double>>
                child 0, item: struct<feature: string, cosine: double>
                    child 0, feature: string
                    child 1, cosine: double
              group_size: int64
              no_real_pii: bool
              to
              {'receipt_id': Value('string'), 'title': Value('string'), 'asset_type': Value('string'), 'asset_source': Value('string'), 'synthetic_only': Value('bool'), 'no_real_pii': Value('bool'), 'no_real_transactions': Value('bool'), 'codec': Value('string'), 'group_size': Value('int64'), 'n_levels': Value('int64'), 'bits_per_weight': Value('int64'), 'tensor_shape': List(Value('int64')), 'n_events': Value('int64'), 'n_features': Value('int64'), 'feature_names': List(Value('string')), 'element_count': Value('int64'), 'dtype': Value('string'), 'seed': Value('int64'), 'fidelity': {'cosine_similarity': Value('float64'), 'cosine_min_per_feature': Value('float64'), 'rms_error': Value('float64'), 'max_abs_error': Value('float64'), 'mse': Value('float64'), 'gate_threshold': Value('float64'), 'gate': Value('string')}, 'per_feature_cosine': List({'feature': Value('string'), 'cosine': Value('float64')}), 'size': {'original_bytes': Value('int64'), 'compressed_bytes': Value('int64'), 'compression_ratio': Value('float64')}, 'hashes': {'sha256_original': Value('string'), 'sha256_decoded': Value('string')}, 'policy_context': {'policy_version': Value('string'), 'adapter_source': Value('string'), 'note': Value('string')}, 'schema_version': Value('string'), 'cost': {'wall_time_s': Value('float64'), 'cpu_time_s': Value('float64'), 'peak_memory_mb': Value('float64'), 'python_version': Value('string'), 'numpy_version': Value('string'), 'hostname': Value('string'), 'timestamp_start': Value('string'), 'timestamp_end': 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.

HXQ Non-LLM Tensor Codec Proof — Regulated Asset Features

This is not a model. This is a tensor-codec proof artifact.

It contains synthetic regulated-asset transfer feature tensors compressed with HXQ, demonstrating that HXQ operates on dense numeric tensors outside LLM weights.

What this proves

HXQ (affine g128, 6-bit) compresses a 16-feature regulated-asset event tensor at 4.92x with cosine similarity 0.9997 and per-feature minimum cosine 0.9997 — all 16 features above the 0.998 gate.

This sits alongside other non-LLM proofs in the HXQ zoo:

Domain	Artifact	Cosine
LLM weights	Qwen, Zamba2, Mamba, TinyLlama, SmolLM3	> 0.999
Hybrid SSM weights	Zamba2, Mamba2	> 0.999
Encoder weights	BERT	> 0.999
Vision embeddings	CLIP	> 0.999
Text embeddings	SBERT (1024 MS MARCO)	0.9996
Regulated-asset features	This artifact	0.9997

What this is NOT

Not a trading model, price predictor, or financial AI
Not an AML/KYC/compliance product
Not a tokenized asset system
Not trained on real transactions
Contains no real PII, wallet addresses, or financial data

All tensor data is synthetic, generated by generate_proof.py with seed=42. The feature distributions are modeled after fields in a regulated-asset risk adapter but contain no real-world information.

File	Description
`tensor_original.npy`	Original synthetic feature tensor (8192×16, float32)
`tensor_decoded.npy`	HXQ-decoded tensor (lossless reproduction)
`hxq_compressed.npz`	Per-feature quantization indices + group params
`receipt.json`	Full fidelity receipt with per-feature cosines and cost
`schema.json`	Feature schema with distributions and ranges
`generate_proof.py`	Tensor generator + HXQ compressor (reproduces everything)
`verify.py`	Independent verification script (exits 0 on PASS)

Features

16 numeric features with realistic distributions:

Feature	Distribution	Per-feature cosine
amount_usd	log-normal	0.9997
velocity_24h	Poisson(5)	0.9999
cumulative_24h_usd	derived	0.9997
risk_score	Beta(2,5)×100	0.9999
counterparty_count_24h	geometric(0.3)	0.9999
kyc_level_encoded	categorical {0,1,2,3}	1.0000
sanctions_score	exponential(5)	0.9998
pep_score	exponential(3)	0.9998
cross_border_flag	Bernoulli(0.25)	1.0000
oracle_risk_signal	Normal(30,20)	0.9999
subthreshold_repeat_count	Poisson(1)	0.9999
time_since_last_tx_hours	exponential(24)	0.9998
wallet_age_days	Uniform(1,3650)	0.9999
jurisdiction_risk_tier	categorical {1,2,3,4}	1.0000
amount_to_cumulative_ratio	derived	0.9999
velocity_acceleration	Normal(0,5)	0.9997

Reproduce

# Generate tensors + compress + receipt
python3 generate_proof.py

# Verify independently
python3 verify.py
# Exit code 0 = PASS

Requires only numpy.

Codec

Method: HXQ affine per-group quantization
Group size: 128
Levels: 64 (6-bit)
Compression: per-feature (each column compressed independently)
Ratio: 4.92x (512 KB → 104 KB)

Provenance

Policy version: regulated_asset_v0.2
Synthetic only: true
No real PII: true
No real transactions: true
Seed: 42
Schema version: 2.0

License

Apache 2.0

Downloads last month: 49

Total file size:

1.23 MB