andreasmartin/apertus-v1.1-swiss-embed-0.4b-bidir
LoRA-style vs swiss-ai/Apertus-v1.1-0.5B. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of k_proj, o_proj, q_proj, v_proj changed, in 20 of 20 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/1, Norms 0/81, MLP 0/120, Attention 80/80
1 other repos hold byte-for-byte identical weight files.