| --- |
| license: other |
| language: |
| - en |
| task_categories: |
| - text-generation |
| tags: |
| - deepshopper |
| - recommendation |
| - fashion |
| - outfit |
| --- |
| |
| # DeepShopper frozen splits (mapper / reward) |
|
|
| Deterministic, leakage-safe train/test splits used across DeepShopper. Split assignment is a |
| stable `sha1(need)` hash (same need never crosses train/test; reproducible). Gender-stratified. |
| Contains: `fashionrec_task1_mapper/{train,test}`, `amz_mapper/{female,male,other}.{train,test}` |
| (need→plan), and `amz_reward_bundle/{female,male,other}.{train,test}` (need→outfit, reward |
| positives). Generated by `scripts/make_mapper_reward_splits.py`. Code: https://github.com/clijo/reco-rl (branch outfit_bundle). |
| |
