training_args.json

{
  "optimizer": "rmsprop",
  "scheduler": "step",
  "learning_rate": 5e-05,
  "batch_size": 24,
  "epochs": 10,
  "seed": 3407,
  "status": "default recipe; not a completed run"
}