README.md

---
license: other
pretty_name: D2F Review Processed Dataset Bundle
tags:
- anonymous-review
- time-series
- processed-datasets
---

# D2F Review Processed Dataset Bundle

This dataset repository hosts a reduced processed-data bundle for anonymous review. It is intended to support subset evaluation using redistributable processed inputs only.

## Files

- `d2f_review_processed_datasets_public.zip`: reduced processed dataset bundle
- `d2f_review_processed_datasets_public.manifest.json`: file list, byte counts, and SHA256 checksums

## Bundle Summary

- Size: 2,336,951,424 bytes (2.18 GiB)
- SHA256: `9b35e971668875f31312eaa4fa0058339a5a1c7ec16aa9a7207c5d2d0d4eb5f2`
- File count: 16
- Uncompressed member bytes: 3,664,989,844

## Included Data

- `data/cryptos_binance_spot_monthly_1s_l10.npz`
- Processed Monash dataset manifests, audits, and `.tsf` source files under `paper_datasets/monash/`

## Excluded Data

The public bundle excludes restricted or non-public-review files:

- `data/es_mbp_10.npz`
- `data/sleep_edf_3ch_100hz_stage_conditioned.npz`
- `data/sleep_edf_3ch_100hz_stage_conditioned.json`
- Monash raw download zip files under `paper_datasets/monash/*/raw/`

## Usage

```python
from huggingface_hub import hf_hub_download

bundle = hf_hub_download(
    repo_id="pixelhero98/d2f-dataset",
    filename="d2f_review_processed_datasets_public.zip",
    repo_type="dataset",
)
```

Then pass the downloaded path to the evaluation runner:

```bash
CUDA_VISIBLE_DEVICES='' PYTHONDONTWRITEBYTECODE=1 python -m diffusion_flow_inference.evaluation.diffusion_flow_time_reparameterization \
  --dataset_bundle_zip "$bundle" \
  --dataset_bundle_mode auto \
  --dataset_bundle_extract_root outputs/dataset_bundles/extracted \
  --backbone_manifest outputs/backbone_matrix/backbone_manifest.json
```

Users are responsible for complying with the terms of the upstream datasets represented in this processed review artifact.