train_results.json

{
    "achieved_tflops_per_gpu": 4.8731586964180895,
    "achieved_tflops_per_gpu_theoretical": 111.80085057316455,
    "epoch": 5.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.3421081006526947,
    "mfu_percent": 0.34439284073626075,
    "mfu_percent_theoretical": 7.901120181849085,
    "total_flos": 1.879814441009152e+18,
    "train_loss": 0.37729120831580676,
    "train_runtime": 24109.2913,
    "train_samples_per_second": 2.073,
    "train_steps_per_second": 0.033,
    "valid_targets_mean": 5900.6,
    "valid_targets_min": 645
}