efficiencyx/Jun-LoRA-12B-GGUF
Full fine-tune vs unsloth/gemma-4-12B-it-qat-q4_0-unquantized. efficiencyx/Jun-LoRA-12B-GGUF is a GGUF (Q8_0) of different weights. Fine-tune: 100% of the weights changed, across 48 of 48 layers.
Tensors changed: Embeddings 1/1, Norms 0/289, MLP 144/144, Attention 184/184