cl-nagoya/ruri-v3-pt-30m
Full fine-tune vs sbintuitions/modernbert-ja-30m. Fine-tune: 100% of the weights changed, across 10 of 10 layers.
Tensors changed: Embeddings 2/2, Norms 20/20, Attention 20/20, MLP 20/20
3 derivatives checked
- cl-nagoya/ruri-v3-30m: Full fine-tune
- hotchpotch/japanese-reranker-xsmall-v2: Full fine-tune
- hotchpotch/japanese-reranker-tiny-v2: Full fine-tune