Power/dpo/train_results.json

{
    "epoch": 9.908256880733944,
    "total_flos": 9.208786937802916e+17,
    "train_loss": 0.6812221950954861,
    "train_runtime": 2021.1055,
    "train_samples_per_second": 8.629,
    "train_steps_per_second": 0.134
}