train_results.json

{
    "achieved_tflops_per_gpu": 115718.58136405371,
    "achieved_tflops_per_gpu_theoretical": 9541873.969677001,
    "epoch": 7.0,
    "mfu_percent": 37089.28889873516,
    "mfu_percent_theoretical": 3058292.9389990387,
    "total_flos": 1.6519984675532308e+18,
    "train_loss": 0.0,
    "train_runtime": 1.7845,
    "train_samples_per_second": 45043.632,
    "train_steps_per_second": 2816.453
}