bias_list[bias]._get_weight(
).data = bias_list[bias]._get_weight(
).data - learning_rate * bias_list[bias]._get_weight().grad.data
bias_list[bias]._get_weight().grad.data.zero_()
if i % 2 == 0:
print("Total Loss : ", total_loss.sum().data)
# Testing
n = 20
x_limit = np.linspace(-2, 2, n)
y_limit = np.linspace(-2, 2, n)
empty_image = np.empty((28 * n, 28 * n))
for i, zi in enumerate(x_limit):
for j, pi in enumerate(y_limit):
generated_latent_layer = np.array([[zi, pi]] * batch_size)
Decoder_Noisy = Decoder(generated_latent_layer)
generated_image = Decoder_Noisy.decode(weight_list, bias_list)
generated_image = generated_image.detach().numpy()
empty_image[(n - i - 1) * 28:(n - i) * 28,
j * 28:(j + 1) * 28] = generated_image[0].reshape(
28, 28)
plt.figure(figsize=(8, 10))
X, Y = np.meshgrid(x_limit, y_limit)
plt.imshow(empty_image, origin="upper", cmap="gray")
plt.grid('False')
plt.show()
