r16_s4_lr5e-5/train_results.json

{
    "epoch": 10.0,
    "total_flos": 48225369194496.0,
    "train_loss": 1.7629533290863038,
    "train_runtime": 1618.6168,
    "train_samples_per_second": 0.612,
    "train_steps_per_second": 0.025
}