Is there any way to gradually increase learning rate using TFOD API?

Question

I am training CenterNet models using Tensorflow Object Detection API. I need to find better learning rate range. I used learning rate finder with Keras models before but I couldn't find any way to implement same strategy with TFOD API. I tried setting up whole training as warm up but somehow learning rate does not start from a low value as much as I desire.

optimizer {
    adam_optimizer {
      learning_rate {
        cosine_decay_learning_rate {
          learning_rate_base: 0.9
          total_steps: 35000
          warmup_learning_rate: 0.00000001
          warmup_steps: 20000
        }
      }
      epsilon: 1e-07
    }
    use_moving_average: false
  }
  num_steps: 20000

When I started the training with the parameters above, Learning rate starts from 0.005. I checked the source code and I couldn't find any restriction.

Is there any way to gradually increase learning rate from such a low value?