LR does not update on solver.steps

[ghazni123]

I am using tools/plain_train_net.py to train model. training logic is exactly same do_train(). My settings are:

cfg.TEST.EVAL_PERIOD = 29544
cfg.SOLVER.IMS_PER_BATCH = 6
cfg.SOLVER.CHECKPOINT_PERIOD = 5000
cfg.SOLVER.BASE_LR = 0.0025
cfg.SOLVER.MAX_ITER = 1000000
cfg.SOLVER.STEPS = (500000, 750000)

Once the step is reached, the training continued with same base_lr and did not update

Do I need to do something additional in order to change the LR at solver.steps?