def build_VAE_prior(self, params):
"""
Build a VAE prior with loaded weights
"""
shape = self.shape
assert self.N == 784
biases, weights = self.load_VAE_prior(params)
if params['id'] == '20_relu_400_sigmoid_784_bias':
D, N1, N = 20, 400, 28 * 28
W1, W2 = weights
b1, b2 = biases
prior_x = (GaussianPrior(size=D) @ V(id="z_0") @ LinearChannel(
W1, name="W_1") @ V(id="Wz_1") @ BiasChannel(b1) @ V(id="b_1")
@ LeakyReluChannel(0) @ V(id="z_1") @ LinearChannel(
W2, name="W_2") @ V(id="Wz_2") @ BiasChannel(b2)
@ V(id="b_2") @ HardTanhChannel() @ V(id="z_2")
@ ReshapeChannel(prev_shape=self.N,
next_shape=self.shape))
else:
raise NotImplementedError
return prior_x
def test_leaky_relu_proba(self):
channel = LeakyReluChannel(slope=0.1)
self._test_function_proba(channel, self.records)
def test_leaky_relu_second_moment(self):
channel = LeakyReluChannel(slope=0.1)
self._test_function_second_moment(channel, self.records, places=2)
def test_leaky_relu_posterior(self):
channel = LeakyReluChannel(slope=0.1)
self._test_function_posterior(channel, self.records, places=6)