b8 stage 2 -- SFR + constant rank r=102, capacity-matched to b6

LoRA rank / alpha: 102 / 204 (scaling: alpha/r)
Full rank schedule: 102 -> 102 -> 102 -> 102 -> 102
Replay (cumulative levels): True
Stage partition: difficulty
Early stopping patience: 2
Cumulative train examples this stage: 1817
Validation split seed: 42 (5% of train, stratified by level; test set never used for selection)
Code commit: c925de05f810a41b16d469627f37f87c9283d7ac

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support