all_results.json

{
    "epoch": 100.0,
    "init_mem_cpu_alloc_delta": 324872,
    "init_mem_cpu_peaked_delta": 18282,
    "init_mem_gpu_alloc_delta": 408400896,
    "init_mem_gpu_peaked_delta": 0,
    "perplexity": 4.663977352453378,
    "train_mem_cpu_alloc_delta": 1622129,
    "train_mem_cpu_peaked_delta": 68939497,
    "train_mem_gpu_alloc_delta": 1641378304,
    "train_mem_gpu_peaked_delta": 6652795904,
    "train_runtime": 40107.6334,
    "train_samples_per_second": 4.48
}