Apple_silicon_benchmark Apple silicon benchmark Nvidia A6000 vs M2 Max 38gpu core 64GB Model:opt-125m(huggingface pytorch) LORA finetune with PEFT FP32 Batchsize = 16 (input sequences less than 30 token) Memory: 7820M vs 32GB GPU-Util: 73% vs 92% Speed: 21it/s vs 7it/s