hotchpotch/japanese-reranker-base-v2
Full fine-tune vs cl-nagoya/ruri-v3-pt-130m. Fine-tune: 100% of the weights changed, across 19 of 19 layers.
Tensors changed: Embeddings 2/2, Norms 38/38, Attention 38/38, MLP 38/38
1 derivatives checked
- hotchpotch/open-provence-reranker-v1: Full fine-tune