ggml-org/Bespoke-Nimble-9B-v3-GGUF
Adapter vs bespokelabs/Bespoke-Nimble-9B-v3. bespokelabs/Bespoke-Nimble-9B-v3 is a LoRA adapter, not full weights: rank 64, on down_proj, gate_proj, in_proj_a, in_proj_b, in_proj_qkv, in_proj_z, k_proj, o_proj… in 32 layers (248 adapted matrices). It modifies ggml-org/Bespoke-Nimble-9B-v3-GGUF at load time.