train_results.json

{
    "achieved_tflops_per_gpu": 10.344710879289135,
    "achieved_tflops_per_gpu_theoretical": 918.1221365250786,
    "epoch": 7.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.11453244835138321,
    "mfu_percent": 3.31561246131062,
    "mfu_percent_theoretical": 294.2699155529098,
    "total_flos": 2.7891230302346936e+18,
    "train_loss": 0.027739192068844568,
    "train_runtime": 33702.2835,
    "train_samples_per_second": 4.331,
    "train_steps_per_second": 0.271,
    "valid_targets_mean": 7108.5,
    "valid_targets_min": 2092
}