amadeusai/Amadeus-Verbo-FI-Qwen2.5-72B-PT-BR-Instruct
Full fine-tune vs Qwen/Qwen2.5-72B-Instruct. Fine-tune: 98% of the weights changed, across 80 of 80 layers.
Tensors changed: Embeddings 0/1, Norms 111/161, MLP 240/240, Attention 559/560, lm_head 1/1