train_results.json

{
    "achieved_tflops_per_gpu": 4.371659502071317,
    "achieved_tflops_per_gpu_theoretical": 731.7873855698042,
    "epoch": 7.0,
    "loss_nan_ranks": 0,
    "loss_rank_avg": 0.12962277233600616,
    "mfu_percent": 0.4420282610790007,
    "mfu_percent_theoretical": 73.99265779270011,
    "total_flos": 2.2298248732265677e+17,
    "train_loss": 0.3448181322210504,
    "train_runtime": 6375.7964,
    "train_samples_per_second": 3.456,
    "train_steps_per_second": 0.216,
    "valid_targets_mean": 2546.5,
    "valid_targets_min": 1046
}