all_results.json

{
    "achieved_tflops_per_gpu": 0.003519985193016927,
    "achieved_tflops_per_gpu_theoretical": 464.14292510677404,
    "epoch": 7.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.17162613570690155,
    "mfu_percent": 0.00024876220445349306,
    "mfu_percent_theoretical": 32.801620148888624,
    "total_flos": 1157609273950208.0,
    "train_loss": 0.22598929736066897,
    "train_runtime": 20554.2284,
    "train_samples_per_second": 2.189,
    "train_steps_per_second": 0.137,
    "valid_targets_mean": 3578.4,
    "valid_targets_min": 1983
}