train_results.json

{
    "achieved_tflops_per_gpu": 2.2154428118318754,
    "achieved_tflops_per_gpu_theoretical": 396.3848338503107,
    "epoch": 5.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.4620312452316284,
    "mfu_percent": 0.15656839659589225,
    "mfu_percent_theoretical": 28.013062462919482,
    "total_flos": 2.3852276139845222e+17,
    "train_loss": 0.5620896111628052,
    "train_runtime": 6728.981,
    "train_samples_per_second": 7.428,
    "train_steps_per_second": 0.117,
    "valid_targets_mean": 702.6,
    "valid_targets_min": 459
}