Intern-S1-mini assay transfer V27 — BBB Martins L3
Validation-selected checkpoint from the V27 BBB Martins L3 record-level assay-transfer run.
Provenance
- Base model:
jiosephlee/Intern-S1-mini-lm - Base model revision:
fcb667c380ae01f57693a45b4b5c2d331052a107 - Training dataset:
jiosephlee/assay-transfer-record-level-v27-bbb-martins-l3-intern - Dataset revision:
2a4c6848de474f0c9152e3acdf4087ce1fe16db8 - Training configuration: 10 scheduled epochs, seed 42, 8 GPUs, device batch size 4, gradient accumulation 4, Flash Attention 2, BFD packing, and padding-free training
- Selection metric: validation
L3/overall/knn_regression_id_ood_level_macro_mae_at_3(lower is better) - Selected optimizer step: 510
- Selected validation metric:
0.5804 - The run was intentionally stopped before completing the full scheduled training; this repository contains its saved validation-selected checkpoint.
- Training run
- Held-out model/Morgan comparison
Held-out test comparison
| Ranker | Binary macro-F1@3 | Binary macro-F1@5 | Regression MAE@3 | Spearman@3 | Top-1 hit@3 |
|---|---|---|---|---|---|
| Model | 0.8611 | 0.8611 | 0.7717 | -0.0802 | 0.1408 |
| Morgan vanilla | 0.7432 | 0.7677 | 1.1246 | -0.0382 | 0.1127 |
| Morgan weighted | 0.7345 | 0.7677 | 1.0652 | -0.1077 | 0.1408 |
Lower regression MAE is better; higher values are better for the other metrics. All W&B metric names retain the L3/ prefix.
The repository contains the full Transformers checkpoint and tokenizer files. metric.json is the complete validation-selection record stored with the checkpoint.