Jaidchen/Focus
Partial fine-tune vs Qwen/Qwen3.6-27B. Partial change: 1% of the weights changed (mlp, norms), in 1 of 64 layers.
Tensors changed: Embeddings 0/1, MLP 3/192, Attention 0/448, lm_head 0/1, Norms 3/209
Partial fine-tune vs Qwen/Qwen3.6-27B. Partial change: 1% of the weights changed (mlp, norms), in 1 of 64 layers.
Tensors changed: Embeddings 0/1, MLP 3/192, Attention 0/448, lm_head 0/1, Norms 3/209