ProCreations/grug-27b-v2

Partial fine-tune vs Qwen/Qwen3.8-27B. Partial change: 89% of the weights changed (attention, mlp, norms, other), in 64 of 64 layers.

Tensors changed: Norms 3/325, Attention 308/560, MLP 195/307, Embeddings 0/5, lm_head 0/1, other 1/1

3 derivatives checked