train_results.json

{
    "achieved_tflops_per_gpu": 4.829163613129045,
    "achieved_tflops_per_gpu_theoretical": 155.84360988986572,
    "epoch": 5.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.13704220950603485,
    "mfu_percent": 0.34128364757095725,
    "mfu_percent_theoretical": 11.013682677729026,
    "total_flos": 4.222945955781018e+16,
    "train_loss": 0.6025812339782715,
    "train_runtime": 546.5421,
    "train_samples_per_second": 2.891,
    "train_steps_per_second": 0.046,
    "valid_targets_mean": 4015.5,
    "valid_targets_min": 740
}