training_log.json

{
  "train_runtime": 665.8011,
  "train_samples_per_second": 1.494,
  "train_steps_per_second": 0.011,
  "total_flos": 1.772375330783232e+16,
  "train_loss": 2.5272955894470215,
  "epoch": 0.9,
  "loss": 2.6658,
  "learning_rate": 0.00021428571428571427
}