all_results.json

{
    "achieved_tflops_per_gpu": 3.9261879574929686,
    "achieved_tflops_per_gpu_theoretical": 155.84360988986572,
    "epoch": 5.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.011615145951509476,
    "mfu_percent": 0.2774691136037434,
    "mfu_percent_theoretical": 11.013682677729026,
    "total_flos": 1.0666132853164605e+18,
    "train_loss": 0.014929962234132609,
    "train_runtime": 16979.149,
    "train_samples_per_second": 2.942,
    "train_steps_per_second": 0.046,
    "valid_targets_mean": 2078.6,
    "valid_targets_min": 397
}