Han-Lei Wang

```javascript learning_rate = ( tf.where( tf.greater_equal(global_step, start_decay_step), tf.train.polynomial_decay(starter_learning_rate, global_step-start_decay_step, decay_steps, end_learning_rate, power=1.0), starter_learning_rate ) ) ``` According to the code above in model.py, if the global_step get around 200000, the...

I have the same problem... it seems that it cannot be built in standalone, and can only work in play &edit mode!?