snwy/frankenqwen3-8B-235B-dense-custom-untuned
Partial fine-tune vs snwy/frankenqwen3-8B-235B-dense-conversion-interleaved-untuned. Partial change: 4% of the weights changed (attention, mlp, norms), in 8 of 64 layers.
Tensors changed: Norms 8/257, MLP 7/192, Attention 20/256, lm_head 0/1, Embeddings 0/1