RefalMachine/ruadapt_qwen2.5_3B_ext_u48_instruct_v4
Adapter vs RefalMachine/ruadapt_qwen2.5_3B_ext_u48_full_lr5e4_peft_mlp_32_32_bs256. RefalMachine/ruadapt_qwen2.5_3B_ext_u48_full_lr5e4_peft_mlp_32_32_bs256 is a LoRA adapter, not full weights: rank 32, on down_proj, gate_proj, k_proj, o_proj, q_proj, up_proj, v_proj in 36 layers (252 adapted matrices). It modifies RefalMachine/ruadapt_qwen2.5_3B_ext_u48_instruct_v4 at load time.
1 derivatives checked
- MaziyarPanahi/ruadapt_qwen2.5_3B_ext_u48_instruct_v4-GGUF: Same weights, converted