r256_s2_lr5e-4/all_results.json

{
    "epoch": 10.0,
    "total_flos": 16141963493376.0,
    "train_loss": 7.470467883348465,
    "train_runtime": 1715.6348,
    "train_samples_per_second": 0.577,
    "train_steps_per_second": 0.023
}