mradermacher/Llama-3.2-1B-FlashNorm-GGUF
LoRA-style vs open-machine/Llama-3.2-1B-FlashNorm-test. mradermacher/Llama-3.2-1B-FlashNorm-GGUF is a GGUF (F16) of different weights. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of gate_proj, k_proj, q_proj, up_proj, v_proj changed, in 16 of 16 layers; norms, biases and embeddings are untouched.
Tensors changed: Norms 0/33, Embeddings 0/1, MLP 32/48, Attention 48/64