NASK-PIB/LLaVA-Bielik-11b-v2.6-instruct
Full fine-tune vs google/siglip2-so400m-patch14-384. Fine-tune: 93% of the weights changed, across 26 of 27 layers.
Tensors changed: Embeddings 3/3, Norms 85/112, MLP 102/112, Attention 208/220, other 0/1
Full fine-tune vs google/siglip2-so400m-patch14-384. Fine-tune: 93% of the weights changed, across 26 of 27 layers.
Tensors changed: Embeddings 3/3, Norms 85/112, MLP 102/112, Attention 208/220, other 0/1