common_params_fit_impl: projected to use 30000 MiB of device memory vs. 30500 MiB of free device memory
common_params_fit: failed to fit params to free device memory: n_gpu_layers already set by user, abort
load_tensors: using device CUDA0 (Synthetic GPU) - 30500 MiB free
load_tensors: offloaded 33/33 layers to GPU
load_tensors: CUDA0 model buffer size = 24000.00 MiB
llama_kv_cache: CUDA0 KV buffer size = 4000.00 MiB
llama_context: CUDA0 compute buffer size = 2000.00 MiB
