epoch: 12 LearningRate: base_lr: 0.01 schedulers: - !PiecewiseDecay gamma: 0.1 milestones: [7, 10] - !LinearWarmup start_factor: 0.3333333333333333 steps: 500 OptimizerBuilder: optimizer: momentum: 0.9 type: Momentum regularizer: factor: 0.0001 type: L2 clip_grad_by_norm: 35
