5CD-AI/Vintern-Embedding-1B
LoRA-style vs 5CD-AI/Vintern-1B-v3_5. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, k_proj, o_proj, q_proj, up_proj, v_proj changed, in 24 of 24 layers; norms, biases and embeddings are untouched.
Tensors changed: lm_head 0/1, Embeddings 0/5, Norms 0/145, MLP 72/174, Attention 96/264, other 0/48
1 other repos hold byte-for-byte identical weight files.