dchen0/font-classifier-v3
LoRA-style vs facebook/dinov2-base. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of query, value changed, in 12 of 12 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/5, Attention 24/96, other 0/24, MLP 0/48, Norms 0/50