nisten/tqwendo-36b

Partial fine-tune vs Qwen/Qwen2.5-Coder-32B-Instruct. Partial change: 89% of the weights changed (attention, mlp, norms), in 63 of 64 layers.

Tensors changed: Embeddings 0/1, Norms 116/129, MLP 179/192, Attention 411/448, lm_head 0/1

3 derivatives checked