train_results.json

{
    "achieved_tflops_per_gpu": 3.5712384827833192,
    "achieved_tflops_per_gpu_theoretical": 382.833215599018,
    "epoch": 7.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.12112488597631454,
    "mfu_percent": 0.36109590321368246,
    "mfu_percent_theoretical": 38.70912190081072,
    "total_flos": 1.1127298177168835e+18,
    "train_loss": 0.27853014678641513,
    "train_runtime": 38947.6166,
    "train_samples_per_second": 1.811,
    "train_steps_per_second": 0.113,
    "valid_targets_mean": 3677.4,
    "valid_targets_min": 1243
}