Intern-S1-mini V27 BBB Martins L2
Full fine-tuned checkpoint selected from the V27 BBB Martins L2 assay-transfer run.
- Base model:
jiosephlee/Intern-S1-mini-lmat revisionfcb667c380ae01f57693a45b4b5c2d331052a107 - Training data:
jiosephlee/assay-transfer-record-level-v27-bbb-martins-l2-internat revision9e04b429a1322a76bc2aadd074f492cfb5cbde65 - Selected checkpoint: step 1575
- Validation selector:
ranking_validation/overall/knn_id_ood_level_macro_mae_f1_error_at_3 - Validation selector value:
0.5000490216050504(lower is better) - Training recipe: 5 epochs configured, batch size 32 per GPU, 4 GPUs, learning rate
2e-5; training was intentionally stopped after preserving this best checkpoint.
metric.json contains the selected checkpoint's validation metrics. The four sharded BF16 model weights, model configuration, and tokenizer files are included.
Validation history
The validation/assay-transfer-record-level-v27-bbb-martins-l2-intern/ folder contains the 76 model validation summaries and matching per-bucket CSVs from step 0 through step 1875. The selected step 1575 metrics are also recorded in metric.json. validation_morgan_vanilla/metrics.json contains the vanilla Morgan aggregate metrics on the same pinned validation_ranking split (521 queries).
| Ranker | Joint ID/OOD error @3 (lower is better) | Spearman | Top-1 hit @3 |
|---|---|---|---|
| Model, step 1575 | 0.5000 | 0.5007 | 0.2376 |
| Vanilla Morgan | 0.7055 | 0.0739 | 0.1881 |
The validation Morgan evaluation is logged in W&B run o26tfw2s. The test comparison is in W&B run hm9qazf5.