athena129/Gemma4Defense-2B
LoRA-style (fewer layers) vs google/gemma-4-E2B-it. athena129/Gemma4Defense-2B matches part of google/gemma-4-E2B-it (600 of its 2011 tensors; the rest of google/gemma-4-E2B-it has no counterpart). On the shared part: Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, k_proj, o_proj, q_proj, up_proj, v_proj changed, in 35 of 35 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/2, Norms 0/247, other 0/106, MLP 105/105, Attention 100/140
1 derivatives checked
- mradermacher/Gemma4Defense-2B-GGUF: Same weights, converted