nllg/detikzify-v2.5-8b
Full fine-tune vs nllg/detikzify-v2-8b. Fine-tune: 95% of the weights changed, across 32 of 32 layers.
Tensors changed: other 1/2, Embeddings 1/4, Norms 13/177, MLP 96/208, Attention 128/348, lm_head 1/1
Full fine-tune vs nllg/detikzify-v2-8b. Fine-tune: 95% of the weights changed, across 32 of 32 layers.
Tensors changed: other 1/2, Embeddings 1/4, Norms 13/177, MLP 96/208, Attention 128/348, lm_head 1/1