b4ph/qwen3-4b-lowram-bench
LoRA-style vs Qwen/Qwen3-4B. b4ph/qwen3-4b-lowram-bench is a GGUF (F16) of different weights. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of up_proj changed, in 1 of 36 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/1, Norms 0/145, MLP 1/108, Attention 0/144