gemma3-1B-ddm/train_results.json

{
    "epoch": 1.0,
    "total_flos": 1.6465434123436032e+19,
    "train_loss": 1.934643552494049,
    "train_runtime": 49285.6778,
    "train_samples_per_second": 155.826,
    "train_steps_per_second": 9.739
}