Atugil-Intelligence/Synas-Detect-1
Full fine-tune (fewer layers) vs openai/clip-vit-large-patch14. Atugil-Intelligence/Synas-Detect-1 matches part of openai/clip-vit-large-patch14 (391 of its 592 tensors; the rest of openai/clip-vit-large-patch14 has no counterpart). On the shared part: Fine-tune: 100% of the weights changed, across 24 of 24 layers.
Tensors changed: Embeddings 3/3, Norms 98/100, MLP 96/96, Attention 192/192