tf.quantization.fake_quant_with_min_max_args_gradient

Compute gradients for a FakeQuantWithMinMaxArgs operation.