train_results.json

{
    "epoch": 30.0,
    "init_mem_cpu_alloc_delta": 20270917,
    "init_mem_cpu_peaked_delta": 18306,
    "init_mem_gpu_alloc_delta": 1261919232,
    "init_mem_gpu_peaked_delta": 0,
    "total_flos": 4.998078272539758e+19,
    "train_mem_cpu_alloc_delta": 57067470,
    "train_mem_cpu_peaked_delta": 487786762,
    "train_mem_gpu_alloc_delta": 3793118208,
    "train_mem_gpu_peaked_delta": 6813593600,
    "train_runtime": 61300.1017,
    "train_samples": 12730,
    "train_samples_per_second": 0.111
}