{
    "before_init_mem_cpu": 19409301504,
    "before_init_mem_gpu": 514785792,
    "epoch": 5.0,
    "eval_accuracy": 0.5020384866275277,
    "eval_loss": 2.396484375,
    "eval_mem_cpu_alloc_delta": 1241088,
    "eval_mem_cpu_peaked_delta": 0,
    "eval_mem_gpu_alloc_delta": 0,
    "eval_mem_gpu_peaked_delta": 13894116352,
    "eval_runtime": 7.6383,
    "eval_samples": 24,
    "eval_samples_per_second": 3.142,
    "eval_steps_per_second": 0.131,
    "init_mem_cpu_alloc_delta": 20480,
    "init_mem_cpu_peaked_delta": 0,
    "init_mem_gpu_alloc_delta": 0,
    "init_mem_gpu_peaked_delta": 0,
    "perplexity": 10.984491067549964,
    "train_loss": 2.6133464177449546,
    "train_mem_cpu_alloc_delta": 87850864640,
    "train_mem_cpu_peaked_delta": 4164583424,
    "train_mem_gpu_alloc_delta": -456804352,
    "train_mem_gpu_peaked_delta": 13894116352,
    "train_runtime": 736.506,
    "train_samples": 165,
    "train_samples_per_second": 1.12,
    "train_steps_per_second": 0.02
}