Created by: Jie-Fang

add auto loss scaling.

Usage:

1. Cast specific inputs to float16

imgs = fluid.layers.cast(images, "float16")

2. Decorate optimizer and minimize scaled loss

optimizer = fluid.optimizer.Adam(learning_rate=0.001)
	
 mp_optimizer = fluid.contrib.mixed_precision.decorate(
	            optimizer=optimizer, init_loss_scaling=8.0, use_dynamic_loss_scaling=True)
	
 scaled_loss, _, _ = mp_optimizer.minimize(avg_cost)

In the decorate function, you can set incr_every__n_steps, decr_every_n_nan_or_inf, incr_ratio, and decr_ratio to control the auto loss scaling details.