train_results.json

{
    "achieved_tflops_per_gpu": 4.5112325337057255,
    "achieved_tflops_per_gpu_theoretical": 247.31703308609124,
    "epoch": 7.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.03830413520336151,
    "mfu_percent": 1.4459078633672198,
    "mfu_percent_theoretical": 79.26827983528565,
    "total_flos": 3.038645528557519e+18,
    "train_loss": 0.05716925004446894,
    "train_runtime": 84196.6554,
    "train_samples_per_second": 1.175,
    "train_steps_per_second": 0.073,
    "valid_targets_mean": 3729.2,
    "valid_targets_min": 1273
}