Lasimeri/GrugCap-27B
LoRA-style vs ProCreations/grug-27b. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, in_proj_a, in_proj_b, in_proj_qkv, in_proj_z, k_proj, o_proj, out_proj, q_proj, up_proj, v_proj changed, in 29 of 64 layers; norms, biases and embeddings are untouched.
Tensors changed: lm_head 0/1, Embeddings 0/4, Norms 0/319, Attention 138/556, MLP 87/304
2 derivatives checked
- mradermacher/GrugCap-27B-i1-GGUF: Quantization
- mradermacher/GrugCap-27B-GGUF: Quantization