vllm-sr/mmbert32k-intent-classifier-merged
LoRA-style vs vllm-sr/mmbert-32k-yarn. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of Wi, Wo, Wqkv changed, in 22 of 22 layers; norms, biases and embeddings are untouched.
Tensors changed: other 0/1, Norms 0/45, Embeddings 0/2, Attention 44/44, MLP 44/44