cl-nagoya/ruri-v3-reranker-310m
Full fine-tune vs cl-nagoya/ruri-v3-pt-310m. Fine-tune: 100% of the weights changed, across 25 of 25 layers.
Tensors changed: Embeddings 2/2, Norms 50/50, Attention 50/50, MLP 50/50
2 derivatives checked
- hotchpotch/open-provence-reranker-large-v1: Full fine-tune
- masahiroid/ruri-v3-reranker-310m-mlx: Same weights, converted
1 other repos hold byte-for-byte identical weight files.