hotchpotch/japanese-reranker-small-v2
Full fine-tune vs cl-nagoya/ruri-v3-pt-70m. Fine-tune: 100% of the weights changed, across 13 of 13 layers.
Tensors changed: Embeddings 2/2, Norms 26/26, Attention 26/26, MLP 26/26
1 derivatives checked
- masahiroid/japanese-reranker-small-v2-mlx: Same weights, converted