avg_train_loss: 0.0
optimizer_step: 65504
step: 0