train_results.json

{
    "epoch": 33.0,
    "init_mem_cpu_alloc_delta": 65159,
    "init_mem_cpu_peaked_delta": 18306,
    "init_mem_gpu_alloc_delta": 1261972480,
    "init_mem_gpu_peaked_delta": 0,
    "total_flos": 4.638603148454345e+19,
    "train_mem_cpu_alloc_delta": 0,
    "train_mem_cpu_peaked_delta": 0,
    "train_mem_gpu_alloc_delta": 6365176832,
    "train_mem_gpu_peaked_delta": 0,
    "train_runtime": 3655.2719,
    "train_samples": 7434,
    "train_samples_per_second": 2.095
}