avg_train_loss: 0.0
optimizer_step: 95128
step: 0