nisten/franqwenstein-35b

Partial fine-tune vs AiCloser/Qwen2.5-32B-AGI. Partial change: 83% of the weights changed (attention, mlp, norms), in 56 of 64 layers.

Tensors changed: Embeddings 0/1, Norms 112/129, MLP 168/192, Attention 392/448, lm_head 0/1

2 derivatives checked