train_results.json

{
    "achieved_tflops_per_gpu": 0.0034246777803463827,
    "achieved_tflops_per_gpu_theoretical": 464.14292510677404,
    "epoch": 7.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.252300500869751,
    "mfu_percent": 0.0002420266982576949,
    "mfu_percent_theoretical": 32.801620148888624,
    "total_flos": 1709579575230464.0,
    "train_loss": 0.21878372598024368,
    "train_runtime": 31199.6428,
    "train_samples_per_second": 2.185,
    "train_steps_per_second": 0.137,
    "valid_targets_mean": 3376.2,
    "valid_targets_min": 180
}