valoomba/RombO1-Fuse

Partial fine-tune vs FuseAI/FuseO1-DeepSeekR1-Qwen2.5-Coder-32B-Preview. Partial change: 34% of the weights changed (attention, lm_head, mlp, norms), in 55 of 64 layers.

Tensors changed: lm_head 1/1, Embeddings 0/1, Norms 16/129, MLP 65/192, Attention 99/448

2 derivatives checked