llama-3-8b-base-beta-dpo-ultrafeedback-4xh200-batch-128-20260424-044124/train_results.json

{
    "epoch": 0.9989528795811519,
    "total_flos": 0.0,
    "train_loss": 4.632088508745909,
    "train_runtime": 6811.5994,
    "train_samples": 61135,
    "train_samples_per_second": 8.975,
    "train_steps_per_second": 0.07
}