PyTorch implementation of NICE: NON-LINEAR INDEPENDENT COMPONENTS ESTIMATION
input [1, 784]
output [1, 784]
4 coupling layer + 1 scaling layer
----------------------------------------------------------------
Layer (type) Output Shape Param #
================================================================
Linear-1 [-1, 1, 1000] 785,000
LeakyReLU-2 [-1, 1, 1000] 0
Linear-3 [-1, 1, 1000] 1,001,000
LeakyReLU-4 [-1, 1, 1000] 0
Linear-5 [-1, 1, 1000] 1,001,000
LeakyReLU-6 [-1, 1, 1000] 0
Linear-7 [-1, 1, 1000] 1,001,000
LeakyReLU-8 [-1, 1, 1000] 0
Linear-9 [-1, 1, 1000] 1,001,000
LeakyReLU-10 [-1, 1, 1000] 0
Linear-11 [-1, 1, 784] 784,784
----------------------------------------------------------------
Linear-12 [-1, 1, 1000] 785,000
LeakyReLU-13 [-1, 1, 1000] 0
Linear-14 [-1, 1, 1000] 1,001,000
LeakyReLU-15 [-1, 1, 1000] 0
Linear-16 [-1, 1, 1000] 1,001,000
LeakyReLU-17 [-1, 1, 1000] 0
Linear-18 [-1, 1, 1000] 1,001,000
LeakyReLU-19 [-1, 1, 1000] 0
Linear-20 [-1, 1, 1000] 1,001,000
LeakyReLU-21 [-1, 1, 1000] 0
Linear-22 [-1, 1, 784] 784,784
----------------------------------------------------------------
Linear-23 [-1, 1, 1000] 785,000
LeakyReLU-24 [-1, 1, 1000] 0
Linear-25 [-1, 1, 1000] 1,001,000
LeakyReLU-26 [-1, 1, 1000] 0
Linear-27 [-1, 1, 1000] 1,001,000
LeakyReLU-28 [-1, 1, 1000] 0
Linear-29 [-1, 1, 1000] 1,001,000
LeakyReLU-30 [-1, 1, 1000] 0
Linear-31 [-1, 1, 1000] 1,001,000
LeakyReLU-32 [-1, 1, 1000] 0
Linear-33 [-1, 1, 784] 784,784
----------------------------------------------------------------
Linear-34 [-1, 1, 1000] 785,000
LeakyReLU-35 [-1, 1, 1000] 0
Linear-36 [-1, 1, 1000] 1,001,000
LeakyReLU-37 [-1, 1, 1000] 0
Linear-38 [-1, 1, 1000] 1,001,000
LeakyReLU-39 [-1, 1, 1000] 0
Linear-40 [-1, 1, 1000] 1,001,000
LeakyReLU-41 [-1, 1, 1000] 0
Linear-42 [-1, 1, 1000] 1,001,000
LeakyReLU-43 [-1, 1, 1000] 0
Linear-44 [-1, 1, 784] 784,784
----------------------------------------------------------------
ScalingLayer-45 [-1, 1, 784] 784
================================================================
Total params: 22,295,920
Trainable params: 22,295,920
Non-trainable params: 0
----------------------------------------------------------------
Input size (MB): 0.002991
Forward/backward pass size (MB): 0.335083
Params size (MB): 85.052185
Estimated Total Size (MB): 85.390259
----------------------------------------------------------------input [1, 784]
output [1, 784]
5 FCN-leakyrelu + 1 FCN(Fully Connection Layer)
----------------------------------------------------------------
Layer (type) Output Shape Param #
================================================================
Linear-1 [-1, 1, 1000] 785,000
LeakyReLU-2 [-1, 1, 1000] 0
Linear-3 [-1, 1, 1000] 1,001,000
LeakyReLU-4 [-1, 1, 1000] 0
Linear-5 [-1, 1, 1000] 1,001,000
LeakyReLU-6 [-1, 1, 1000] 0
Linear-7 [-1, 1, 1000] 1,001,000
LeakyReLU-8 [-1, 1, 1000] 0
Linear-9 [-1, 1, 1000] 1,001,000
LeakyReLU-10 [-1, 1, 1000] 0
Linear-11 [-1, 1, 784] 784,784
================================================================
Total params: 5,573,784
Trainable params: 5,573,784
Non-trainable params: 0
----------------------------------------------------------------
Input size (MB): 0.002991
Forward/backward pass size (MB): 0.082275
Params size (MB): 21.262299
Estimated Total Size (MB): 21.347565
----------------------------------------------------------------from torch.utils.tensorboard import SummaryWriter
# default dir ./runs
writer = SummaryWriter()
# run
$ tensorboard --logdir ./runs