zentropi-ai/cope-b-a4b
LoRA-style (fewer layers) vs google/gemma-4-26B-A4B-it. zentropi-ai/cope-b-a4b matches part of google/gemma-4-26B-A4B-it (657 of its 1013 tensors; the rest of google/gemma-4-26B-A4B-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 k_proj, o_proj, q_proj, v_proj changed, in 30 of 30 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/1, MLP 0/150, Norms 0/271, other 0/120, Attention 115/115
1 derivatives checked
- mradermacher/cope-b-a4b-GGUF: Quantization