| --- |
| license: cc-by-4.0 |
| pretty_name: CoPeP Reviewer Sample |
| task_categories: |
| - other |
| tags: |
| - protein |
| - biology |
| - continual-learning |
| - parquet |
| configs: |
| - config_name: default |
| data_files: |
| - split: train |
| path: train/train_sample.parquet |
| - split: validation |
| path: val/val.parquet |
| --- |
| |
| # CoPeP Reviewer Sample |
|
|
| This repository contains a compact deterministic sample of `anon2435/CoPeP` for |
| reviewer inspection. It is intended for checking file format, schema, and example |
| records without downloading the full dataset. |
|
|
| ## Contents |
|
|
| - `train/train_sample.parquet`: deterministic stratified train-row sample. |
| - `val/val.parquet`: full validation parquet from the anonymous CoPeP release. |
| - `dataset_manifest.json`: sampling details, row counts, and limitations. |
|
|
| The train sample preserves the main dataset train-row schema: |
| `name`, `sequence`, `seq_hash`, `task_0` through `task_9`, and `row_idx`. |
|
|
| ## Sampling |
|
|
| Rows are grouped by first observed temporal task, defined as the first true |
| `task_0` through `task_9` column. Up to 10,000 rows are selected per group using |
| a deterministic score derived from `seq_hash`, `row_idx`, and a fixed seed. |
|
|
| This sample is not a drop-in replacement for the full dataset. It does not |
| include task split files, because those files reference row indices in the full |
| training table. |
|
|
