Phonepadith/aidc-5k-laos-gemma-3-4b-it-v2
LoRA-style vs google/gemma-3-4b-it. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of q_proj, v_proj changed, in 19 of 34 layers; norms, biases and embeddings are untouched.
Tensors changed: Norms 0/316, MLP 0/210, Attention 68/352, other 0/1, Embeddings 0/3
2 derivatives checked
- mradermacher/aidc-5k-laos-gemma-3-4b-it-v2-i1-GGUF: Quantization (fewer layers)
- mradermacher/aidc-5k-laos-gemma-3-4b-it-v2-GGUF: Same weights, converted (fewer layers)