record Keras models created in the context of a probmodel

[rcabanasdepaz]

Consider the following model:

@inf.probmodel
def vae(k, d0, d, decoder):
    with inf.datamodel():
        z = inf.Normal(tf.ones(k), 1,name="z")
        x = inf.Normal(decoder(d0, d, z), 1, name="x")

def decoder(d0, d, z):
    return inf.keras.Sequential([
        tf.keras.layers.Dense(d0),
        tf.keras.layers.Dense(d, activation="relu"),
    ], name="decoder")(z)



p = vae(k=2, d0=100, d=28*28, decoder=decoder)

We might need to access to the keras model defined:

>>> p.prior().keras()
 {'decoder': <tensorflow.python.keras.engine.sequential.Sequential at 0x12d722668>}

>>> p.posterior().keras()
 {'decoder': <tensorflow.python.keras.engine.sequential.Sequential at 0x12e842634>}

Like with RV, these could be the prior or posterior models.