llama_model_loader: - tensor split  0:                token_embd.weight f16      [   4096,  16384,      1,      1 ]   128.00 MiB
0.02.934.812 I load_tensors: offloaded 33/33 layers to GPU
0.02.934.818 I load_tensors:          CPU model buffer size =   128.00 MiB
0.02.934.819 I load_tensors:        CUDA0 model buffer size =  4096.00 MiB
0.02.945.000 I llama_context:        CUDA0 compute buffer size =    64.00 MiB
