mnoukhov/nuevamol-360m-reinvent-sft
Partial fine-tune vs mnoukhov/nuevamol-360M-6Btok-wd5. Partial change: 49% of the weights changed (attention, embeddings, mlp, norms), in 32 of 32 layers.
Tensors changed: Embeddings 1/1, Norms 51/65, MLP 34/96, Attention 112/128