aozalevsky/search_slm_demo
Partial fine-tune vs Qwen/Qwen3.5-4B. Partial change: 77% of the weights changed (attention, mlp, norms), in 32 of 32 layers.
Tensors changed: Embeddings 0/5, MLP 96/199, Attention 104/324, Norms 24/209, other 0/1
Partial fine-tune vs Qwen/Qwen3.5-4B. Partial change: 77% of the weights changed (attention, mlp, norms), in 32 of 32 layers.
Tensors changed: Embeddings 0/5, MLP 96/199, Attention 104/324, Norms 24/209, other 0/1