README.md

---
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/*.parquet
---

# PrismLayersObjectTransparency

![object_transparency_edit_pipeline](https://cdn-uploads.huggingface.co/production/uploads/681f211189050843b62d602d/b5pFPTON-PCdU1xNA0nX1.png)

This dataset repo stores the transparency dataset as parquet shards.

## Contents

- `data/*.parquet`: training split shards
- `metadata/`: upload manifest and integrity reports

## Upload Selection

- Selected samples uploaded in this run: 5679
- Shard count: 89
- Samples per shard: 64
- Parquet row group size: 1
- Parquet page index: True
- Source manifest: `uploadable_paths.json`
- Total uploadable samples in manifest: 5679

## Load

```python
from datasets import load_dataset

ds = load_dataset("artplus/PrismLayersObjectTransparency", split="train")
```