First of all, tensorflow2.0 does not have a GELU activation function, so the GELU activation function must be customized.
Secondly, the custom code is as follows:
Step one: import tensorflow as tf
Step 2: class GELU(layers.Layer): def __init__(self): super(GELU, self).__init__() def call(self, x): cdf = 0.5 * (1.0 + tf.tanh((np.sqrt(2 / np.pi) * (x + 0.044715 * tf.pow(x, 3))))) return x * cdf
third step:
self.act=GRLU()
Step 4: (used in call)
x = self.act(x)