mergekit-community/MS3-RP-RP-half2

Full fine-tune vs Nohobby/MS3-test-Merge-1. Fine-tune: 100% of the weights changed, across 40 of 40 layers.

Tensors changed: lm_head 1/1, Embeddings 1/1, Norms 6/81, MLP 120/120, Attention 160/160

2 derivatives checked