euclaise/Memphis-scribe-3B

Partial fine-tune vs euclaise/Memphis-CoT-3B. Partial change: 89% of the weights changed (attention, mlp, norms), in 32 of 32 layers.

Tensors changed: lm_head 0/1, Embeddings 0/1, Norms 127/130, MLP 96/96, Attention 121/128

1 derivatives checked