使用Pytorch实现变分自编码器

最新推荐文章于 2024-04-25 00:03:39 发布
最新推荐文章于 2024-04-25 00:03:39 发布
阅读量446 收藏 1
点赞数
分类专栏： Pytorch教程 文章标签： pytorch 人工智能 python 自编码器 神经网络
版权声明：本文为博主原创文章，遵循 CC 4.0 BY-SA 版权协议，转载请附上原文出处链接和本声明。
本文链接：https://blog.csdn.net/AIHUBEI/article/details/134892330
版权
使用Pytorch实现变分自编码器

可以结合这篇VAE讲解文章阅读这篇blog post代码。

# Import necessary packages.
import os
import torch
import torch.nn as nn
import torch.nn.functional as F
import torchvision
from torchvision import transforms
from torchvision.utils import save_image

# Device configuration.
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
print(torch.cuda.get_device_name())

NVIDIA GeForce GTX 1650

# Create a directory if not exists.
sample_dir = 'samples'
if not os.path.exists(sample_dir):
    os.makedirs(sample_dir)

# Hyper-parameters.
image_size = 784
h_dim = 400
z_dim = 20
num_epochs = 15
batch_size = 128
learn_rate = 1e-3

# Load the MNIST dataset.
dataset = torchvision.datasets.MNIST(root='../../data/',
                                     train=True,
                                     transform=transforms.ToTensor(),
                                     download=True
                                     )

# Define the Data Loader.
data_loader = torch.utils.data.DataLoader(dataset=dataset,
                                          batch_size=batch_size,
                                          shuffle=True)

# Devine the VAE model.
class VAE(nn.Module):
    def __init__(self, image_size=784, h_dim=400, z_dim=20):
        super(VAE, self).__init__()
        self.fc1 = nn.Linear(image_size, h_dim)
        self.fc2 = nn.Linear(h_dim, z_dim)
        self.fc3 = nn.Linear(h_dim, z_dim)
        self.fc4 = nn.Linear(z_dim, h_dim)
        self.fc5 = nn.Linear(h_dim, image_size)
    
    # define the encoder.
    def encode(self, x):
        h = F.relu(self.fc1(x))
        return self.fc2(h), self.fc3(h)
    
    # 重参数
    def reparameterize(self, mu, log_var):
        std = torch.exp(log_var/2)
        eps = torch.randn_like(std)
        return mu + eps * std
    
    # define the decoder.
    def decode(self, z):
        h = F.relu(self.fc4(z))
        return F.sigmoid(self.fc5(h))
    
    # forward.
    def forward(self, x):
        mu, log_var = self.encode(x)
        z = self.reparameterize(mu, log_var)
        x_reconst = self.decode(z)
        return x_reconst, mu, log_var

model = VAE().to(device)

# define the optimizer.
optimizer = torch.optim.Adam(model.parameters(), lr=learn_rate)

# Start training.
for epoch in range(num_epochs):
    for i, (x, _) in enumerate(data_loader):
        # Forward pass
        x = x.to(device).view(-1, image_size)
        x_reconst, mu, log_var = model(x)

        # Compute reconstruction loss and kl divergence.
        # For KL divergence, see Appendix B in VAE paper or http://yunjey47.tistory.com/43
        reconst_loss = F.binary_cross_entropy(x_reconst, x, size_average=False)
        kl_div = - 0.5 * torch.sum(1 + log_var - mu.pow(2) - log_var.exp())

        # Backprop and optimize.
        loss = reconst_loss + kl_div
        optimizer.zero_grad()
        loss.backward()
        optimizer.step()

        # Set an output countor.
        if (i+1) % 10 == 0:
            print('Epoch[{}/{}], Step [{}/{}], Reconst Loss: {:.4f}, KL Div: {:.4f}'
                  .format(epoch+1, num_epochs, i+1, len(data_loader), reconst_loss.item(), kl_div.item()))
    
    with torch.no_grad():
        # Save the sample images.
        z = torch.randn(batch_size, z_dim).to(device)
        out = model.decode(z).view(-1, 1, 28, 28)
        save_image(out, os.path.join(sample_dir, 'sample-{}.png'.format(epoch+1)))

        # Save the reconstructed images.
        out, _, _ = model(x)
        x_reconst = torch.cat([x.view(-1, 1, 28, 28), out.view(-1, 1, 28, 28)], dim=3)
        save_image(x_reconst, os.path.join(sample_dir, 'reconst-{}.png'.format(epoch+1)))

/home/wsl_ubuntu/anaconda3/envs/xy_trans/lib/python3.8/site-packages/torch/nn/_reduction.py:42: UserWarning: size_average and reduce args will be deprecated, please use reduction='sum' instead.
  warnings.warn(warning.format(ret))


Epoch[1/15], Step [10/469], Reconst Loss: 36616.9141, KL Div: 3740.0823
Epoch[1/15], Step [20/469], Reconst Loss: 29773.6797, KL Div: 975.3514
Epoch[1/15], Step [30/469], Reconst Loss: 27105.4824, KL Div: 1213.4338
Epoch[1/15], Step [40/469], Reconst Loss: 26462.3633, KL Div: 635.6080
Epoch[1/15], Step [50/469], Reconst Loss: 25826.6094, KL Div: 663.0458
Epoch[1/15], Step [60/469], Reconst Loss: 25826.2461, KL Div: 789.0715
Epoch[1/15], Step [70/469], Reconst Loss: 24164.5078, KL Div: 921.9595
Epoch[1/15], Step [80/469], Reconst Loss: 24587.4492, KL Div: 1041.5199
Epoch[1/15], Step [90/469], Reconst Loss: 22938.7578, KL Div: 1195.8701
Epoch[1/15], Step [100/469], Reconst Loss: 22055.6328, KL Div: 1312.9821
Epoch[1/15], Step [110/469], Reconst Loss: 21499.7246, KL Div: 1558.9216
Epoch[1/15], Step [120/469], Reconst Loss: 19639.1953, KL Div: 1837.0837
Epoch[1/15], Step [130/469], Reconst Loss: 19362.3203, KL Div: 1719.7855
Epoch[1/15], Step [140/469], Reconst Loss: 18975.9766, KL Div: 1942.4675
Epoch[1/15], Step [150/469], Reconst Loss: 18912.7930, KL Div: 1858.0500
Epoch[1/15], Step [160/469], Reconst Loss: 18639.9473, KL Div: 1861.7341
Epoch[1/15], Step [170/469], Reconst Loss: 18130.4434, KL Div: 1787.8860
Epoch[1/15], Step [180/469], Reconst Loss: 18130.5273, KL Div: 2044.2058
Epoch[1/15], Step [190/469], Reconst Loss: 17782.9336, KL Div: 1964.3074
Epoch[1/15], Step [200/469], Reconst Loss: 17909.9258, KL Div: 1933.2725
Epoch[1/15], Step [210/469], Reconst Loss: 16536.2812, KL Div: 2063.9639
Epoch[1/15], Step [220/469], Reconst Loss: 17814.0332, KL Div: 2057.8086
Epoch[1/15], Step [230/469], Reconst Loss: 17407.0820, KL Div: 2108.3965
Epoch[1/15], Step [240/469], Reconst Loss: 16596.4531, KL Div: 2106.2700
Epoch[1/15], Step [250/469], Reconst Loss: 16938.6289, KL Div: 2231.6538
Epoch[1/15], Step [260/469], Reconst Loss: 16497.2539, KL Div: 2105.1924
Epoch[1/15], Step [270/469], Reconst Loss: 15581.2100, KL Div: 2217.1523
Epoch[1/15], Step [280/469], Reconst Loss: 16118.3672, KL Div: 2281.6221
Epoch[1/15], Step [290/469], Reconst Loss: 15486.8379, KL Div: 2341.7874
Epoch[1/15], Step [300/469], Reconst Loss: 16615.8828, KL Div: 2368.1182
Epoch[1/15], Step [310/469], Reconst Loss: 15342.2422, KL Div: 2366.4150
Epoch[1/15], Step [320/469], Reconst Loss: 15661.7598, KL Div: 2309.4153
Epoch[1/15], Step [330/469], Reconst Loss: 15456.1113, KL Div: 2324.4585
Epoch[1/15], Step [340/469], Reconst Loss: 15182.8945, KL Div: 2390.0042
Epoch[1/15], Step [350/469], Reconst Loss: 15453.9336, KL Div: 2552.0754
Epoch[1/15], Step [360/469], Reconst Loss: 15737.4033, KL Div: 2574.0132
Epoch[1/15], Step [370/469], Reconst Loss: 15173.9482, KL Div: 2714.6523
Epoch[1/15], Step [380/469], Reconst Loss: 14593.1543, KL Div: 2465.7563
Epoch[1/15], Step [390/469], Reconst Loss: 14496.3486, KL Div: 2471.1484
Epoch[1/15], Step [400/469], Reconst Loss: 15301.4873, KL Div: 2655.8091
Epoch[1/15], Step [410/469], Reconst Loss: 14106.9092, KL Div: 2595.3081
Epoch[1/15], Step [420/469], Reconst Loss: 13720.7979, KL Div: 2626.6914
Epoch[1/15], Step [430/469], Reconst Loss: 14111.7471, KL Div: 2679.8662
Epoch[1/15], Step [440/469], Reconst Loss: 14438.3672, KL Div: 2592.5044
Epoch[1/15], Step [450/469], Reconst Loss: 13193.9785, KL Div: 2707.9358
Epoch[1/15], Step [460/469], Reconst Loss: 14041.0977, KL Div: 2559.0266
Epoch[2/15], Step [10/469], Reconst Loss: 13620.3125, KL Div: 2716.8257
Epoch[2/15], Step [20/469], Reconst Loss: 13833.0693, KL Div: 2752.4922
Epoch[2/15], Step [30/469], Reconst Loss: 13145.0605, KL Div: 2830.2148
Epoch[2/15], Step [40/469], Reconst Loss: 13201.6924, KL Div: 2690.5637
Epoch[2/15], Step [50/469], Reconst Loss: 13244.9805, KL Div: 2750.5166
Epoch[2/15], Step [60/469], Reconst Loss: 13621.4131, KL Div: 2782.6863
Epoch[2/15], Step [70/469], Reconst Loss: 13467.2559, KL Div: 2725.6943
Epoch[2/15], Step [80/469], Reconst Loss: 13888.3320, KL Div: 2820.3740
Epoch[2/15], Step [90/469], Reconst Loss: 13611.9551, KL Div: 2758.0264
Epoch[2/15], Step [100/469], Reconst Loss: 13340.2197, KL Div: 2882.6914
Epoch[2/15], Step [110/469], Reconst Loss: 13055.8193, KL Div: 2964.9287
Epoch[2/15], Step [120/469], Reconst Loss: 13338.4805, KL Div: 2900.1157
Epoch[2/15], Step [130/469], Reconst Loss: 12345.2256, KL Div: 2874.3140
Epoch[2/15], Step [140/469], Reconst Loss: 13200.5859, KL Div: 2832.7090
Epoch[2/15], Step [150/469], Reconst Loss: 12849.0508, KL Div: 2767.7114
Epoch[2/15], Step [160/469], Reconst Loss: 13169.3955, KL Div: 2982.5615
Epoch[2/15], Step [170/469], Reconst Loss: 13079.1592, KL Div: 2904.1270
Epoch[2/15], Step [180/469], Reconst Loss: 12614.8232, KL Div: 2821.8394
Epoch[2/15], Step [190/469], Reconst Loss: 12650.0430, KL Div: 2924.8147
Epoch[2/15], Step [200/469], Reconst Loss: 12591.2227, KL Div: 2943.8188
Epoch[2/15], Step [210/469], Reconst Loss: 12165.0840, KL Div: 2887.9551
Epoch[2/15], Step [220/469], Reconst Loss: 12595.1963, KL Div: 2930.0762
Epoch[2/15], Step [230/469], Reconst Loss: 12445.6689, KL Div: 2921.4336
Epoch[2/15], Step [240/469], Reconst Loss: 12692.6660, KL Div: 2989.3516
Epoch[2/15], Step [250/469], Reconst Loss: 12472.9375, KL Div: 3079.8313
Epoch[2/15], Step [260/469], Reconst Loss: 12889.5625, KL Div: 2909.4849
Epoch[2/15], Step [270/469], Reconst Loss: 12022.7754, KL Div: 2885.7964
Epoch[2/15], Step [280/469], Reconst Loss: 12365.2314, KL Div: 2887.3723
Epoch[2/15], Step [290/469], Reconst Loss: 12088.4805, KL Div: 2919.9077
Epoch[2/15], Step [300/469], Reconst Loss: 11855.0879, KL Div: 3038.1257
Epoch[2/15], Step [310/469], Reconst Loss: 12560.5947, KL Div: 2883.4067
Epoch[2/15], Step [320/469], Reconst Loss: 11956.1133, KL Div: 2991.7896
Epoch[2/15], Step [330/469], Reconst Loss: 12317.2031, KL Div: 2924.3845
Epoch[2/15], Step [340/469], Reconst Loss: 12212.5273, KL Div: 3035.3462
Epoch[2/15], Step [350/469], Reconst Loss: 12334.0469, KL Div: 3019.7427
Epoch[2/15], Step [360/469], Reconst Loss: 12050.0527, KL Div: 3079.3311
Epoch[2/15], Step [370/469], Reconst Loss: 11681.4121, KL Div: 2893.3408
Epoch[2/15], Step [380/469], Reconst Loss: 13003.4150, KL Div: 3099.3091
Epoch[2/15], Step [390/469], Reconst Loss: 11820.4883, KL Div: 2967.7417
Epoch[2/15], Step [400/469], Reconst Loss: 12099.2207, KL Div: 3034.6660
Epoch[2/15], Step [410/469], Reconst Loss: 12138.3145, KL Div: 3041.2932
Epoch[2/15], Step [420/469], Reconst Loss: 12329.2314, KL Div: 3047.4082
Epoch[2/15], Step [430/469], Reconst Loss: 11969.9053, KL Div: 2886.3508
Epoch[2/15], Step [440/469], Reconst Loss: 12390.2568, KL Div: 3109.6245
Epoch[2/15], Step [450/469], Reconst Loss: 12120.2139, KL Div: 2969.8428
Epoch[2/15], Step [460/469], Reconst Loss: 11519.6641, KL Div: 3061.2007
Epoch[3/15], Step [10/469], Reconst Loss: 11936.5957, KL Div: 2964.1772
Epoch[3/15], Step [20/469], Reconst Loss: 11800.0039, KL Div: 2978.1863
Epoch[3/15], Step [30/469], Reconst Loss: 11555.3711, KL Div: 2970.8296
Epoch[3/15], Step [40/469], Reconst Loss: 11624.9033, KL Div: 3148.7576
Epoch[3/15], Step [50/469], Reconst Loss: 11968.5918, KL Div: 3108.2646
Epoch[3/15], Step [60/469], Reconst Loss: 12142.9434, KL Div: 2936.3755
Epoch[3/15], Step [70/469], Reconst Loss: 11720.2529, KL Div: 3075.7925
Epoch[3/15], Step [80/469], Reconst Loss: 12009.3633, KL Div: 3025.7983
Epoch[3/15], Step [90/469], Reconst Loss: 11264.7637, KL Div: 2961.6201
Epoch[3/15], Step [100/469], Reconst Loss: 11657.8027, KL Div: 3210.4092
Epoch[3/15], Step [110/469], Reconst Loss: 11906.7383, KL Div: 2999.9087
Epoch[3/15], Step [120/469], Reconst Loss: 11462.5273, KL Div: 3038.7141
Epoch[3/15], Step [130/469], Reconst Loss: 11444.2969, KL Div: 3084.1289
Epoch[3/15], Step [140/469], Reconst Loss: 11838.1748, KL Div: 3135.4673
Epoch[3/15], Step [150/469], Reconst Loss: 11309.9902, KL Div: 3102.3538
Epoch[3/15], Step [160/469], Reconst Loss: 11665.4375, KL Div: 3129.4873
Epoch[3/15], Step [170/469], Reconst Loss: 11889.5469, KL Div: 3134.8652
Epoch[3/15], Step [180/469], Reconst Loss: 11875.4521, KL Div: 3102.3916
Epoch[3/15], Step [190/469], Reconst Loss: 11439.8613, KL Div: 3043.8538
Epoch[3/15], Step [200/469], Reconst Loss: 10776.7715, KL Div: 3075.7236
Epoch[3/15], Step [210/469], Reconst Loss: 11236.3750, KL Div: 3091.6584
Epoch[3/15], Step [220/469], Reconst Loss: 11596.7461, KL Div: 3196.4287
Epoch[3/15], Step [230/469], Reconst Loss: 12306.6367, KL Div: 3172.8916
Epoch[3/15], Step [240/469], Reconst Loss: 11366.2051, KL Div: 3156.8931
Epoch[3/15], Step [250/469], Reconst Loss: 11073.8887, KL Div: 3051.8367
Epoch[3/15], Step [260/469], Reconst Loss: 11458.4199, KL Div: 3104.0569
Epoch[3/15], Step [270/469], Reconst Loss: 11336.3613, KL Div: 3068.1055
Epoch[3/15], Step [280/469], Reconst Loss: 11858.4570, KL Div: 3093.3188
Epoch[3/15], Step [290/469], Reconst Loss: 11563.0645, KL Div: 3206.7769
Epoch[3/15], Step [300/469], Reconst Loss: 11524.7051, KL Div: 3143.6553
Epoch[3/15], Step [310/469], Reconst Loss: 11095.1602, KL Div: 3078.8447
Epoch[3/15], Step [320/469], Reconst Loss: 11435.6738, KL Div: 3125.8306
Epoch[3/15], Step [330/469], Reconst Loss: 11313.6299, KL Div: 3047.7759
Epoch[3/15], Step [340/469], Reconst Loss: 11694.4961, KL Div: 3085.8320
Epoch[3/15], Step [350/469], Reconst Loss: 10993.4961, KL Div: 3107.8262
Epoch[3/15], Step [360/469], Reconst Loss: 11378.9541, KL Div: 3155.8989
Epoch[3/15], Step [370/469], Reconst Loss: 10975.0156, KL Div: 3118.0667
Epoch[3/15], Step [380/469], Reconst Loss: 11323.3516, KL Div: 3320.9897
Epoch[3/15], Step [390/469], Reconst Loss: 11334.2871, KL Div: 3027.3093
Epoch[3/15], Step [400/469], Reconst Loss: 11099.5430, KL Div: 3088.1921
Epoch[3/15], Step [410/469], Reconst Loss: 11220.4287, KL Div: 3087.4292
Epoch[3/15], Step [420/469], Reconst Loss: 10945.1504, KL Div: 3013.8545
Epoch[3/15], Step [430/469], Reconst Loss: 11275.8730, KL Div: 3135.9106
Epoch[3/15], Step [440/469], Reconst Loss: 11453.5879, KL Div: 3100.9600
Epoch[3/15], Step [450/469], Reconst Loss: 11319.0762, KL Div: 3123.6206
Epoch[3/15], Step [460/469], Reconst Loss: 11407.2090, KL Div: 3193.6418
Epoch[4/15], Step [10/469], Reconst Loss: 11718.5234, KL Div: 3280.7061
Epoch[4/15], Step [20/469], Reconst Loss: 11384.2910, KL Div: 3232.1904
Epoch[4/15], Step [30/469], Reconst Loss: 11241.3145, KL Div: 3071.8535
Epoch[4/15], Step [40/469], Reconst Loss: 11590.3398, KL Div: 3278.6108
Epoch[4/15], Step [50/469], Reconst Loss: 10744.5977, KL Div: 3076.8062
Epoch[4/15], Step [60/469], Reconst Loss: 11124.6836, KL Div: 3106.5952
Epoch[4/15], Step [70/469], Reconst Loss: 10991.4355, KL Div: 3190.8896
Epoch[4/15], Step [80/469], Reconst Loss: 11381.7734, KL Div: 3210.6592
Epoch[4/15], Step [90/469], Reconst Loss: 11081.2568, KL Div: 3187.8215
Epoch[4/15], Step [100/469], Reconst Loss: 11215.9980, KL Div: 3133.1924
Epoch[4/15], Step [110/469], Reconst Loss: 11682.7969, KL Div: 3114.1912
Epoch[4/15], Step [120/469], Reconst Loss: 11027.0684, KL Div: 3305.7222
Epoch[4/15], Step [130/469], Reconst Loss: 11924.5801, KL Div: 3055.8384
Epoch[4/15], Step [140/469], Reconst Loss: 11066.5186, KL Div: 3323.8628
Epoch[4/15], Step [150/469], Reconst Loss: 11265.1006, KL Div: 3147.9014
Epoch[4/15], Step [160/469], Reconst Loss: 11057.5137, KL Div: 3200.9863
Epoch[4/15], Step [170/469], Reconst Loss: 11322.3828, KL Div: 3177.7881
Epoch[4/15], Step [180/469], Reconst Loss: 11383.3877, KL Div: 3196.4248
Epoch[4/15], Step [190/469], Reconst Loss: 11155.8496, KL Div: 3159.9341
Epoch[4/15], Step [200/469], Reconst Loss: 11340.9756, KL Div: 3213.3047
Epoch[4/15], Step [210/469], Reconst Loss: 11284.0410, KL Div: 3209.7910
Epoch[4/15], Step [220/469], Reconst Loss: 11212.4189, KL Div: 3142.0354
Epoch[4/15], Step [230/469], Reconst Loss: 11293.5273, KL Div: 3137.4514
Epoch[4/15], Step [240/469], Reconst Loss: 11118.7383, KL Div: 3134.6753
Epoch[4/15], Step [250/469], Reconst Loss: 11266.2871, KL Div: 3206.3857
Epoch[4/15], Step [260/469], Reconst Loss: 11535.9268, KL Div: 3239.4302
Epoch[4/15], Step [270/469], Reconst Loss: 11310.2090, KL Div: 3230.6040
Epoch[4/15], Step [280/469], Reconst Loss: 11058.5039, KL Div: 3128.0918
Epoch[4/15], Step [290/469], Reconst Loss: 11320.7031, KL Div: 3206.9297
Epoch[4/15], Step [300/469], Reconst Loss: 11003.5762, KL Div: 3232.1726
Epoch[4/15], Step [310/469], Reconst Loss: 10903.9902, KL Div: 3178.8320
Epoch[4/15], Step [320/469], Reconst Loss: 11044.6514, KL Div: 3210.5181
Epoch[4/15], Step [330/469], Reconst Loss: 11566.7451, KL Div: 3146.3750
Epoch[4/15], Step [340/469], Reconst Loss: 11291.4971, KL Div: 3237.8911
Epoch[4/15], Step [350/469], Reconst Loss: 11393.1807, KL Div: 3211.4863
Epoch[4/15], Step [360/469], Reconst Loss: 11014.6289, KL Div: 3185.3413
Epoch[4/15], Step [370/469], Reconst Loss: 11069.9414, KL Div: 3051.4692
Epoch[4/15], Step [380/469], Reconst Loss: 11403.6025, KL Div: 3289.3984
Epoch[4/15], Step [390/469], Reconst Loss: 11195.8242, KL Div: 3116.9404
Epoch[4/15], Step [400/469], Reconst Loss: 11109.6582, KL Div: 3200.8376
Epoch[4/15], Step [410/469], Reconst Loss: 11192.2852, KL Div: 3221.7708
Epoch[4/15], Step [420/469], Reconst Loss: 11117.5352, KL Div: 3156.1177
Epoch[4/15], Step [430/469], Reconst Loss: 10460.2891, KL Div: 3070.9087
Epoch[4/15], Step [440/469], Reconst Loss: 10897.3438, KL Div: 3154.3621
Epoch[4/15], Step [450/469], Reconst Loss: 10343.1992, KL Div: 3069.1011
Epoch[4/15], Step [460/469], Reconst Loss: 10866.3574, KL Div: 3209.5576
Epoch[5/15], Step [10/469], Reconst Loss: 11480.0293, KL Div: 3253.2832
Epoch[5/15], Step [20/469], Reconst Loss: 10760.3984, KL Div: 3284.6387
Epoch[5/15], Step [30/469], Reconst Loss: 10960.1895, KL Div: 3139.7441
Epoch[5/15], Step [40/469], Reconst Loss: 11209.6641, KL Div: 3130.0898
Epoch[5/15], Step [50/469], Reconst Loss: 11632.1553, KL Div: 3199.1096
Epoch[5/15], Step [60/469], Reconst Loss: 10749.2109, KL Div: 3290.9502
Epoch[5/15], Step [70/469], Reconst Loss: 11001.4668, KL Div: 3180.0503
Epoch[5/15], Step [80/469], Reconst Loss: 11122.5449, KL Div: 3209.7297
Epoch[5/15], Step [90/469], Reconst Loss: 11061.4551, KL Div: 3175.9792
Epoch[5/15], Step [100/469], Reconst Loss: 10843.1787, KL Div: 3233.2974
Epoch[5/15], Step [110/469], Reconst Loss: 10777.2500, KL Div: 3237.2104
Epoch[5/15], Step [120/469], Reconst Loss: 11261.6592, KL Div: 3149.5342
Epoch[5/15], Step [130/469], Reconst Loss: 11682.9150, KL Div: 3226.8076
Epoch[5/15], Step [140/469], Reconst Loss: 10805.0527, KL Div: 3311.6199
Epoch[5/15], Step [150/469], Reconst Loss: 11067.4648, KL Div: 3264.8203
Epoch[5/15], Step [160/469], Reconst Loss: 11352.6855, KL Div: 3211.7300
Epoch[5/15], Step [170/469], Reconst Loss: 11213.7803, KL Div: 3198.3235
Epoch[5/15], Step [180/469], Reconst Loss: 10904.0723, KL Div: 3136.8364
Epoch[5/15], Step [190/469], Reconst Loss: 11041.4141, KL Div: 3146.3555
Epoch[5/15], Step [200/469], Reconst Loss: 11130.6670, KL Div: 3171.1870
Epoch[5/15], Step [210/469], Reconst Loss: 10444.7500, KL Div: 3255.3108
Epoch[5/15], Step [220/469], Reconst Loss: 11035.2383, KL Div: 3164.2109
Epoch[5/15], Step [230/469], Reconst Loss: 10684.4805, KL Div: 3097.4158
Epoch[5/15], Step [240/469], Reconst Loss: 10984.2705, KL Div: 3253.7698
Epoch[5/15], Step [250/469], Reconst Loss: 10203.0811, KL Div: 3159.4170
Epoch[5/15], Step [260/469], Reconst Loss: 10921.9287, KL Div: 3142.5562
Epoch[5/15], Step [270/469], Reconst Loss: 10854.9141, KL Div: 3187.5186
Epoch[5/15], Step [280/469], Reconst Loss: 11198.7490, KL Div: 3142.6655
Epoch[5/15], Step [290/469], Reconst Loss: 10405.9004, KL Div: 3163.6431
Epoch[5/15], Step [300/469], Reconst Loss: 10949.9443, KL Div: 3213.2842
Epoch[5/15], Step [310/469], Reconst Loss: 10765.2363, KL Div: 3084.9729
Epoch[5/15], Step [320/469], Reconst Loss: 11036.6035, KL Div: 3230.5505
Epoch[5/15], Step [330/469], Reconst Loss: 10910.4170, KL Div: 3276.6008
Epoch[5/15], Step [340/469], Reconst Loss: 11100.9385, KL Div: 3233.5229
Epoch[5/15], Step [350/469], Reconst Loss: 10757.4463, KL Div: 3188.2432
Epoch[5/15], Step [360/469], Reconst Loss: 10551.0918, KL Div: 3201.9780
Epoch[5/15], Step [370/469], Reconst Loss: 10979.7646, KL Div: 3148.2729
Epoch[5/15], Step [380/469], Reconst Loss: 10873.8457, KL Div: 3190.8938
Epoch[5/15], Step [390/469], Reconst Loss: 10719.6416, KL Div: 3152.4390
Epoch[5/15], Step [400/469], Reconst Loss: 10831.0283, KL Div: 3280.8018
Epoch[5/15], Step [410/469], Reconst Loss: 10744.4414, KL Div: 3183.4048
Epoch[5/15], Step [420/469], Reconst Loss: 10661.5996, KL Div: 3195.2507
Epoch[5/15], Step [430/469], Reconst Loss: 10606.8096, KL Div: 3155.7129
Epoch[5/15], Step [440/469], Reconst Loss: 10266.9805, KL Div: 3145.1309
Epoch[5/15], Step [450/469], Reconst Loss: 10697.7949, KL Div: 3306.5874
Epoch[5/15], Step [460/469], Reconst Loss: 10815.0723, KL Div: 3061.5244
Epoch[6/15], Step [10/469], Reconst Loss: 10750.5186, KL Div: 3091.4343
Epoch[6/15], Step [20/469], Reconst Loss: 10767.5156, KL Div: 3187.1230
Epoch[6/15], Step [30/469], Reconst Loss: 10963.2441, KL Div: 3315.2053
Epoch[6/15], Step [40/469], Reconst Loss: 11402.0723, KL Div: 3183.5483
Epoch[6/15], Step [50/469], Reconst Loss: 10817.4199, KL Div: 3227.4023
Epoch[6/15], Step [60/469], Reconst Loss: 10604.6230, KL Div: 3223.0859
Epoch[6/15], Step [70/469], Reconst Loss: 10482.5723, KL Div: 3136.7629
Epoch[6/15], Step [80/469], Reconst Loss: 10381.5547, KL Div: 3229.3076
Epoch[6/15], Step [90/469], Reconst Loss: 10782.4814, KL Div: 3115.7874
Epoch[6/15], Step [100/469], Reconst Loss: 10853.3516, KL Div: 3206.6821
Epoch[6/15], Step [110/469], Reconst Loss: 11187.8594, KL Div: 3233.2971
Epoch[6/15], Step [120/469], Reconst Loss: 10842.4863, KL Div: 3170.5535
Epoch[6/15], Step [130/469], Reconst Loss: 10651.2285, KL Div: 3164.5842
Epoch[6/15], Step [140/469], Reconst Loss: 10928.7861, KL Div: 3257.0200
Epoch[6/15], Step [150/469], Reconst Loss: 10734.4785, KL Div: 3161.8726
Epoch[6/15], Step [160/469], Reconst Loss: 11339.3066, KL Div: 3271.0107
Epoch[6/15], Step [170/469], Reconst Loss: 10852.5449, KL Div: 3174.1211
Epoch[6/15], Step [180/469], Reconst Loss: 10806.9805, KL Div: 3301.3025
Epoch[6/15], Step [190/469], Reconst Loss: 10688.3838, KL Div: 3111.3535
Epoch[6/15], Step [200/469], Reconst Loss: 10767.8203, KL Div: 3238.5581
Epoch[6/15], Step [210/469], Reconst Loss: 10700.6094, KL Div: 3260.3083
Epoch[6/15], Step [220/469], Reconst Loss: 10570.7383, KL Div: 3153.7571
Epoch[6/15], Step [230/469], Reconst Loss: 10500.3398, KL Div: 3198.9004
Epoch[6/15], Step [240/469], Reconst Loss: 10538.9473, KL Div: 3140.9490
Epoch[6/15], Step [250/469], Reconst Loss: 10707.0234, KL Div: 3160.1277
Epoch[6/15], Step [260/469], Reconst Loss: 10768.5127, KL Div: 3251.2061
Epoch[6/15], Step [270/469], Reconst Loss: 10869.8545, KL Div: 3130.7891
Epoch[6/15], Step [280/469], Reconst Loss: 10487.0332, KL Div: 3383.9670
Epoch[6/15], Step [290/469], Reconst Loss: 10755.4492, KL Div: 3143.0466
Epoch[6/15], Step [300/469], Reconst Loss: 11032.5059, KL Div: 3187.1768
Epoch[6/15], Step [310/469], Reconst Loss: 10592.7334, KL Div: 3156.4619
Epoch[6/15], Step [320/469], Reconst Loss: 10994.3047, KL Div: 3293.8271
Epoch[6/15], Step [330/469], Reconst Loss: 10492.3770, KL Div: 3268.3318
Epoch[6/15], Step [340/469], Reconst Loss: 10874.0693, KL Div: 3213.5630
Epoch[6/15], Step [350/469], Reconst Loss: 10822.4336, KL Div: 3177.2644
Epoch[6/15], Step [360/469], Reconst Loss: 11128.2539, KL Div: 3267.9502
Epoch[6/15], Step [370/469], Reconst Loss: 10538.5957, KL Div: 3191.8623
Epoch[6/15], Step [380/469], Reconst Loss: 10708.6680, KL Div: 3110.2969
Epoch[6/15], Step [390/469], Reconst Loss: 10597.6621, KL Div: 3220.0962
Epoch[6/15], Step [400/469], Reconst Loss: 10775.5996, KL Div: 3050.2397
Epoch[6/15], Step [410/469], Reconst Loss: 10507.6465, KL Div: 3243.5991
Epoch[6/15], Step [420/469], Reconst Loss: 10911.7246, KL Div: 3177.3521
Epoch[6/15], Step [430/469], Reconst Loss: 10863.1387, KL Div: 3133.7156
Epoch[6/15], Step [440/469], Reconst Loss: 10646.1699, KL Div: 3330.3628
Epoch[6/15], Step [450/469], Reconst Loss: 10464.0215, KL Div: 3142.5591
Epoch[6/15], Step [460/469], Reconst Loss: 10707.5957, KL Div: 3225.0454
Epoch[7/15], Step [10/469], Reconst Loss: 10632.2939, KL Div: 3291.3938
Epoch[7/15], Step [20/469], Reconst Loss: 10564.2461, KL Div: 3145.9785
Epoch[7/15], Step [30/469], Reconst Loss: 11033.7441, KL Div: 3239.8870
Epoch[7/15], Step [40/469], Reconst Loss: 10435.2979, KL Div: 3213.6694
Epoch[7/15], Step [50/469], Reconst Loss: 10564.0449, KL Div: 3205.0608
Epoch[7/15], Step [60/469], Reconst Loss: 10644.0605, KL Div: 3379.3101
Epoch[7/15], Step [70/469], Reconst Loss: 10556.4902, KL Div: 3134.3267
Epoch[7/15], Step [80/469], Reconst Loss: 10741.8203, KL Div: 3274.3567
Epoch[7/15], Step [90/469], Reconst Loss: 10343.9502, KL Div: 3112.3765
Epoch[7/15], Step [100/469], Reconst Loss: 10495.3262, KL Div: 3284.4951
Epoch[7/15], Step [110/469], Reconst Loss: 10845.0107, KL Div: 3286.0674
Epoch[7/15], Step [120/469], Reconst Loss: 10187.9238, KL Div: 3110.1138
Epoch[7/15], Step [130/469], Reconst Loss: 10058.6816, KL Div: 3183.2754
Epoch[7/15], Step [140/469], Reconst Loss: 11230.1758, KL Div: 3380.3442
Epoch[7/15], Step [150/469], Reconst Loss: 10640.2695, KL Div: 3025.5276
Epoch[7/15], Step [160/469], Reconst Loss: 10565.2725, KL Div: 3219.5183
Epoch[7/15], Step [170/469], Reconst Loss: 10699.0156, KL Div: 3276.2739
Epoch[7/15], Step [180/469], Reconst Loss: 10708.7988, KL Div: 3280.9546
Epoch[7/15], Step [190/469], Reconst Loss: 10057.8047, KL Div: 3156.7397
Epoch[7/15], Step [200/469], Reconst Loss: 10406.2402, KL Div: 3208.5554
Epoch[7/15], Step [210/469], Reconst Loss: 10224.6797, KL Div: 3243.7461
Epoch[7/15], Step [220/469], Reconst Loss: 11324.8223, KL Div: 3251.8716
Epoch[7/15], Step [230/469], Reconst Loss: 10306.9883, KL Div: 3193.4456
Epoch[7/15], Step [240/469], Reconst Loss: 10637.6348, KL Div: 3223.8604
Epoch[7/15], Step [250/469], Reconst Loss: 10435.1318, KL Div: 3253.7817
Epoch[7/15], Step [260/469], Reconst Loss: 10495.0957, KL Div: 3175.7251
Epoch[7/15], Step [270/469], Reconst Loss: 10537.0811, KL Div: 3325.2441
Epoch[7/15], Step [280/469], Reconst Loss: 10602.4043, KL Div: 3133.4087
Epoch[7/15], Step [290/469], Reconst Loss: 10619.4512, KL Div: 3237.4973
Epoch[7/15], Step [300/469], Reconst Loss: 11278.1426, KL Div: 3288.4448
Epoch[7/15], Step [310/469], Reconst Loss: 10482.0605, KL Div: 3117.4426
Epoch[7/15], Step [320/469], Reconst Loss: 10585.0801, KL Div: 3187.9077
Epoch[7/15], Step [330/469], Reconst Loss: 10699.5273, KL Div: 3266.8779
Epoch[7/15], Step [340/469], Reconst Loss: 10207.9219, KL Div: 3136.3557
Epoch[7/15], Step [350/469], Reconst Loss: 9954.6777, KL Div: 3164.3784
Epoch[7/15], Step [360/469], Reconst Loss: 10057.3262, KL Div: 3137.8042
Epoch[7/15], Step [370/469], Reconst Loss: 10427.5879, KL Div: 3268.8921
Epoch[7/15], Step [380/469], Reconst Loss: 11133.9805, KL Div: 3244.0298
Epoch[7/15], Step [390/469], Reconst Loss: 10745.1523, KL Div: 3357.5542
Epoch[7/15], Step [400/469], Reconst Loss: 10222.7441, KL Div: 3226.5781
Epoch[7/15], Step [410/469], Reconst Loss: 10608.1641, KL Div: 3265.4609
Epoch[7/15], Step [420/469], Reconst Loss: 10226.9512, KL Div: 3064.1594
Epoch[7/15], Step [430/469], Reconst Loss: 10613.5264, KL Div: 3273.1016
Epoch[7/15], Step [440/469], Reconst Loss: 11280.3701, KL Div: 3273.0532
Epoch[7/15], Step [450/469], Reconst Loss: 10453.7852, KL Div: 3309.5806
Epoch[7/15], Step [460/469], Reconst Loss: 10734.3438, KL Div: 3255.2981
Epoch[8/15], Step [10/469], Reconst Loss: 10222.9199, KL Div: 3162.2825
Epoch[8/15], Step [20/469], Reconst Loss: 10011.3857, KL Div: 3290.9229
Epoch[8/15], Step [30/469], Reconst Loss: 10320.5039, KL Div: 3266.3218
Epoch[8/15], Step [40/469], Reconst Loss: 10001.5527, KL Div: 3128.0627
Epoch[8/15], Step [50/469], Reconst Loss: 10245.0957, KL Div: 3185.4368
Epoch[8/15], Step [60/469], Reconst Loss: 10446.2246, KL Div: 3247.0127
Epoch[8/15], Step [70/469], Reconst Loss: 10673.6357, KL Div: 3304.0286
Epoch[8/15], Step [80/469], Reconst Loss: 10669.7314, KL Div: 3189.3696
Epoch[8/15], Step [90/469], Reconst Loss: 10385.3584, KL Div: 3212.2327
Epoch[8/15], Step [100/469], Reconst Loss: 10694.5918, KL Div: 3266.3623
Epoch[8/15], Step [110/469], Reconst Loss: 10549.9199, KL Div: 3290.3311
Epoch[8/15], Step [120/469], Reconst Loss: 10919.4326, KL Div: 3275.3154
Epoch[8/15], Step [130/469], Reconst Loss: 10440.5811, KL Div: 3252.5164
Epoch[8/15], Step [140/469], Reconst Loss: 10981.5742, KL Div: 3278.2183
Epoch[8/15], Step [150/469], Reconst Loss: 10981.8848, KL Div: 3280.6777
Epoch[8/15], Step [160/469], Reconst Loss: 10529.8359, KL Div: 3185.7778
Epoch[8/15], Step [170/469], Reconst Loss: 10671.3975, KL Div: 3207.3862
Epoch[8/15], Step [180/469], Reconst Loss: 10648.8613, KL Div: 3257.8528
Epoch[8/15], Step [190/469], Reconst Loss: 10642.4365, KL Div: 3295.2463
Epoch[8/15], Step [200/469], Reconst Loss: 10848.7617, KL Div: 3363.5200
Epoch[8/15], Step [210/469], Reconst Loss: 10579.8320, KL Div: 3176.8962
Epoch[8/15], Step [220/469], Reconst Loss: 10305.0020, KL Div: 3200.8408
Epoch[8/15], Step [230/469], Reconst Loss: 10259.6992, KL Div: 3216.2588
Epoch[8/15], Step [240/469], Reconst Loss: 10210.6631, KL Div: 3158.8115
Epoch[8/15], Step [250/469], Reconst Loss: 10581.1133, KL Div: 3098.4463
Epoch[8/15], Step [260/469], Reconst Loss: 10548.2168, KL Div: 3215.6663
Epoch[8/15], Step [270/469], Reconst Loss: 10532.7578, KL Div: 3263.5186
Epoch[8/15], Step [280/469], Reconst Loss: 10147.6152, KL Div: 3119.3459
Epoch[8/15], Step [290/469], Reconst Loss: 10594.5195, KL Div: 3292.4692
Epoch[8/15], Step [300/469], Reconst Loss: 10922.3145, KL Div: 3330.0107
Epoch[8/15], Step [310/469], Reconst Loss: 10477.1426, KL Div: 3282.3530
Epoch[8/15], Step [320/469], Reconst Loss: 10249.3867, KL Div: 3326.6125
Epoch[8/15], Step [330/469], Reconst Loss: 10236.3457, KL Div: 3217.5508
Epoch[8/15], Step [340/469], Reconst Loss: 10678.8818, KL Div: 3249.8584
Epoch[8/15], Step [350/469], Reconst Loss: 10571.1328, KL Div: 3218.7478
Epoch[8/15], Step [360/469], Reconst Loss: 10866.8477, KL Div: 3267.1755
Epoch[8/15], Step [370/469], Reconst Loss: 10645.7363, KL Div: 3110.9197
Epoch[8/15], Step [380/469], Reconst Loss: 10385.8838, KL Div: 3174.8872
Epoch[8/15], Step [390/469], Reconst Loss: 10650.8047, KL Div: 3213.5557
Epoch[8/15], Step [400/469], Reconst Loss: 10668.5947, KL Div: 3285.7410
Epoch[8/15], Step [410/469], Reconst Loss: 10591.2236, KL Div: 3222.0774
Epoch[8/15], Step [420/469], Reconst Loss: 10296.6787, KL Div: 3215.7456
Epoch[8/15], Step [430/469], Reconst Loss: 10314.2236, KL Div: 3188.1904
Epoch[8/15], Step [440/469], Reconst Loss: 10467.1367, KL Div: 3241.2246
Epoch[8/15], Step [450/469], Reconst Loss: 10422.1367, KL Div: 3248.3264
Epoch[8/15], Step [460/469], Reconst Loss: 10665.0215, KL Div: 3266.1738
Epoch[9/15], Step [10/469], Reconst Loss: 10196.8691, KL Div: 3178.3267
Epoch[9/15], Step [20/469], Reconst Loss: 10098.2773, KL Div: 3135.8237
Epoch[9/15], Step [30/469], Reconst Loss: 10363.6641, KL Div: 3224.7319
Epoch[9/15], Step [40/469], Reconst Loss: 10555.9258, KL Div: 3179.8088
Epoch[9/15], Step [50/469], Reconst Loss: 10878.3027, KL Div: 3282.8516
Epoch[9/15], Step [60/469], Reconst Loss: 10450.6465, KL Div: 3172.1494
Epoch[9/15], Step [70/469], Reconst Loss: 10261.3652, KL Div: 3217.4451
Epoch[9/15], Step [80/469], Reconst Loss: 10151.3027, KL Div: 3225.1514
Epoch[9/15], Step [90/469], Reconst Loss: 10402.9609, KL Div: 3146.9385
Epoch[9/15], Step [100/469], Reconst Loss: 10744.1914, KL Div: 3264.0083
Epoch[9/15], Step [110/469], Reconst Loss: 10059.4062, KL Div: 3043.5938
Epoch[9/15], Step [120/469], Reconst Loss: 10452.4492, KL Div: 3272.4497
Epoch[9/15], Step [130/469], Reconst Loss: 10109.1318, KL Div: 3111.4502
Epoch[9/15], Step [140/469], Reconst Loss: 10361.4355, KL Div: 3193.3979
Epoch[9/15], Step [150/469], Reconst Loss: 10786.0176, KL Div: 3227.8223
Epoch[9/15], Step [160/469], Reconst Loss: 10772.4268, KL Div: 3175.7617
Epoch[9/15], Step [170/469], Reconst Loss: 10644.6094, KL Div: 3376.1594
Epoch[9/15], Step [180/469], Reconst Loss: 10415.0000, KL Div: 3223.3853
Epoch[9/15], Step [190/469], Reconst Loss: 10796.9951, KL Div: 3289.1919
Epoch[9/15], Step [200/469], Reconst Loss: 10783.0713, KL Div: 3260.4243
Epoch[9/15], Step [210/469], Reconst Loss: 10893.0830, KL Div: 3373.7468
Epoch[9/15], Step [220/469], Reconst Loss: 10836.8115, KL Div: 3204.9941
Epoch[9/15], Step [230/469], Reconst Loss: 10456.2773, KL Div: 3275.3662
Epoch[9/15], Step [240/469], Reconst Loss: 10204.2227, KL Div: 3193.3169
Epoch[9/15], Step [250/469], Reconst Loss: 10149.0850, KL Div: 3222.0654
Epoch[9/15], Step [260/469], Reconst Loss: 10439.5312, KL Div: 3267.2422
Epoch[9/15], Step [270/469], Reconst Loss: 10568.4404, KL Div: 3184.8958
Epoch[9/15], Step [280/469], Reconst Loss: 10789.6836, KL Div: 3138.2075
Epoch[9/15], Step [290/469], Reconst Loss: 10422.8340, KL Div: 3260.5234
Epoch[9/15], Step [300/469], Reconst Loss: 10229.7598, KL Div: 3207.5962
Epoch[9/15], Step [310/469], Reconst Loss: 10687.4180, KL Div: 3161.0713
Epoch[9/15], Step [320/469], Reconst Loss: 10733.6777, KL Div: 3452.2300
Epoch[9/15], Step [330/469], Reconst Loss: 10646.9961, KL Div: 3243.2581
Epoch[9/15], Step [340/469], Reconst Loss: 9829.2861, KL Div: 3151.9353
Epoch[9/15], Step [350/469], Reconst Loss: 10592.4092, KL Div: 3316.3369
Epoch[9/15], Step [360/469], Reconst Loss: 10463.0449, KL Div: 3251.4734
Epoch[9/15], Step [370/469], Reconst Loss: 10664.1172, KL Div: 3201.3977
Epoch[9/15], Step [380/469], Reconst Loss: 10219.7070, KL Div: 3243.8945
Epoch[9/15], Step [390/469], Reconst Loss: 10187.8633, KL Div: 3214.5801
Epoch[9/15], Step [400/469], Reconst Loss: 9953.3008, KL Div: 3193.5020
Epoch[9/15], Step [410/469], Reconst Loss: 10499.8320, KL Div: 3195.8511
Epoch[9/15], Step [420/469], Reconst Loss: 10737.6045, KL Div: 3270.7095
Epoch[9/15], Step [430/469], Reconst Loss: 10496.1660, KL Div: 3289.7112
Epoch[9/15], Step [440/469], Reconst Loss: 10412.8184, KL Div: 3193.3955
Epoch[9/15], Step [450/469], Reconst Loss: 10529.5479, KL Div: 3377.8003
Epoch[9/15], Step [460/469], Reconst Loss: 10154.2305, KL Div: 3239.6172
Epoch[10/15], Step [10/469], Reconst Loss: 10236.5410, KL Div: 3213.1616
Epoch[10/15], Step [20/469], Reconst Loss: 10612.2803, KL Div: 3262.2549
Epoch[10/15], Step [30/469], Reconst Loss: 10679.3330, KL Div: 3247.2385
Epoch[10/15], Step [40/469], Reconst Loss: 10614.4746, KL Div: 3249.0859
Epoch[10/15], Step [50/469], Reconst Loss: 10106.0898, KL Div: 3193.5972
Epoch[10/15], Step [60/469], Reconst Loss: 10295.7402, KL Div: 3189.0188
Epoch[10/15], Step [70/469], Reconst Loss: 10105.2402, KL Div: 3285.5776
Epoch[10/15], Step [80/469], Reconst Loss: 10795.5781, KL Div: 3159.6721
Epoch[10/15], Step [90/469], Reconst Loss: 10204.5869, KL Div: 3290.6047
Epoch[10/15], Step [100/469], Reconst Loss: 10498.1865, KL Div: 3360.2505
Epoch[10/15], Step [110/469], Reconst Loss: 10509.6484, KL Div: 3240.7234
Epoch[10/15], Step [120/469], Reconst Loss: 10551.1484, KL Div: 3296.1765
Epoch[10/15], Step [130/469], Reconst Loss: 10696.8984, KL Div: 3312.8765
Epoch[10/15], Step [140/469], Reconst Loss: 9975.9434, KL Div: 3204.0149
Epoch[10/15], Step [150/469], Reconst Loss: 10749.1562, KL Div: 3280.9580
Epoch[10/15], Step [160/469], Reconst Loss: 10558.5303, KL Div: 3276.9092
Epoch[10/15], Step [170/469], Reconst Loss: 10976.7295, KL Div: 3353.9478
Epoch[10/15], Step [180/469], Reconst Loss: 10085.6924, KL Div: 3207.1685
Epoch[10/15], Step [190/469], Reconst Loss: 10486.7246, KL Div: 3159.7339
Epoch[10/15], Step [200/469], Reconst Loss: 10099.4775, KL Div: 3174.8127
Epoch[10/15], Step [210/469], Reconst Loss: 10209.8496, KL Div: 3233.4775
Epoch[10/15], Step [220/469], Reconst Loss: 10603.9912, KL Div: 3309.9707
Epoch[10/15], Step [230/469], Reconst Loss: 10298.6387, KL Div: 3270.9316
Epoch[10/15], Step [240/469], Reconst Loss: 10173.3301, KL Div: 3170.9163
Epoch[10/15], Step [250/469], Reconst Loss: 10026.7910, KL Div: 3236.4868
Epoch[10/15], Step [260/469], Reconst Loss: 10553.7070, KL Div: 3179.6345
Epoch[10/15], Step [270/469], Reconst Loss: 10175.8242, KL Div: 3330.2834
Epoch[10/15], Step [280/469], Reconst Loss: 10591.2900, KL Div: 3219.5210
Epoch[10/15], Step [290/469], Reconst Loss: 10519.7139, KL Div: 3270.1125
Epoch[10/15], Step [300/469], Reconst Loss: 10444.9531, KL Div: 3203.0942
Epoch[10/15], Step [310/469], Reconst Loss: 10812.2871, KL Div: 3274.4155
Epoch[10/15], Step [320/469], Reconst Loss: 10295.7109, KL Div: 3139.6843
Epoch[10/15], Step [330/469], Reconst Loss: 10694.4297, KL Div: 3165.6001
Epoch[10/15], Step [340/469], Reconst Loss: 10554.6348, KL Div: 3321.5742
Epoch[10/15], Step [350/469], Reconst Loss: 10688.0059, KL Div: 3252.3789
Epoch[10/15], Step [360/469], Reconst Loss: 10486.1543, KL Div: 3178.7944
Epoch[10/15], Step [370/469], Reconst Loss: 10239.1787, KL Div: 3316.7739
Epoch[10/15], Step [380/469], Reconst Loss: 10374.2188, KL Div: 3132.5322
Epoch[10/15], Step [390/469], Reconst Loss: 10329.8125, KL Div: 3208.3450
Epoch[10/15], Step [400/469], Reconst Loss: 10409.2012, KL Div: 3332.0391
Epoch[10/15], Step [410/469], Reconst Loss: 10529.1855, KL Div: 3274.3516
Epoch[10/15], Step [420/469], Reconst Loss: 11063.1924, KL Div: 3268.7520
Epoch[10/15], Step [430/469], Reconst Loss: 10256.4453, KL Div: 3231.0925
Epoch[10/15], Step [440/469], Reconst Loss: 10324.7969, KL Div: 3219.4368
Epoch[10/15], Step [450/469], Reconst Loss: 10317.6650, KL Div: 3239.9331
Epoch[10/15], Step [460/469], Reconst Loss: 10360.7959, KL Div: 3189.5708
Epoch[11/15], Step [10/469], Reconst Loss: 10053.4727, KL Div: 3250.8550
Epoch[11/15], Step [20/469], Reconst Loss: 10236.5088, KL Div: 3255.9810
Epoch[11/15], Step [30/469], Reconst Loss: 10176.1611, KL Div: 3121.9722
Epoch[11/15], Step [40/469], Reconst Loss: 10437.1982, KL Div: 3250.6987
Epoch[11/15], Step [50/469], Reconst Loss: 10708.3945, KL Div: 3278.7578
Epoch[11/15], Step [60/469], Reconst Loss: 10547.8115, KL Div: 3267.8232
Epoch[11/15], Step [70/469], Reconst Loss: 10649.3711, KL Div: 3313.7485
Epoch[11/15], Step [80/469], Reconst Loss: 10084.4355, KL Div: 3141.8008
Epoch[11/15], Step [90/469], Reconst Loss: 10295.0977, KL Div: 3230.8088
Epoch[11/15], Step [100/469], Reconst Loss: 10348.7480, KL Div: 3272.9365
Epoch[11/15], Step [110/469], Reconst Loss: 10151.3496, KL Div: 3310.7529
Epoch[11/15], Step [120/469], Reconst Loss: 10194.9375, KL Div: 3173.0044
Epoch[11/15], Step [130/469], Reconst Loss: 10214.9697, KL Div: 3293.3711
Epoch[11/15], Step [140/469], Reconst Loss: 10298.1816, KL Div: 3218.0227
Epoch[11/15], Step [150/469], Reconst Loss: 10152.0449, KL Div: 3367.3867
Epoch[11/15], Step [160/469], Reconst Loss: 10273.2090, KL Div: 3232.4277
Epoch[11/15], Step [170/469], Reconst Loss: 10385.1719, KL Div: 3256.7419
Epoch[11/15], Step [180/469], Reconst Loss: 10500.4834, KL Div: 3294.9968
Epoch[11/15], Step [190/469], Reconst Loss: 10245.0488, KL Div: 3300.9277
Epoch[11/15], Step [200/469], Reconst Loss: 10301.4082, KL Div: 3216.1733
Epoch[11/15], Step [210/469], Reconst Loss: 10016.0576, KL Div: 3248.4077
Epoch[11/15], Step [220/469], Reconst Loss: 10252.1484, KL Div: 3350.2583
Epoch[11/15], Step [230/469], Reconst Loss: 10751.9082, KL Div: 3199.2019
Epoch[11/15], Step [240/469], Reconst Loss: 10531.6982, KL Div: 3364.4631
Epoch[11/15], Step [250/469], Reconst Loss: 10466.8740, KL Div: 3298.8354
Epoch[11/15], Step [260/469], Reconst Loss: 10577.1758, KL Div: 3266.2664
Epoch[11/15], Step [270/469], Reconst Loss: 10609.9297, KL Div: 3205.9143
Epoch[11/15], Step [280/469], Reconst Loss: 10435.7598, KL Div: 3367.1514
Epoch[11/15], Step [290/469], Reconst Loss: 10704.1250, KL Div: 3236.8464
Epoch[11/15], Step [300/469], Reconst Loss: 10193.8691, KL Div: 3176.0522
Epoch[11/15], Step [310/469], Reconst Loss: 10056.8320, KL Div: 3177.2244
Epoch[11/15], Step [320/469], Reconst Loss: 10633.1270, KL Div: 3276.2898
Epoch[11/15], Step [330/469], Reconst Loss: 9991.3809, KL Div: 3107.3315
Epoch[11/15], Step [340/469], Reconst Loss: 10530.6885, KL Div: 3171.1392
Epoch[11/15], Step [350/469], Reconst Loss: 10466.3594, KL Div: 3269.9668
Epoch[11/15], Step [360/469], Reconst Loss: 10428.1621, KL Div: 3304.3184
Epoch[11/15], Step [370/469], Reconst Loss: 10165.8691, KL Div: 3153.9817
Epoch[11/15], Step [380/469], Reconst Loss: 10641.2578, KL Div: 3152.3586
Epoch[11/15], Step [390/469], Reconst Loss: 10268.7070, KL Div: 3264.2173
Epoch[11/15], Step [400/469], Reconst Loss: 10582.3369, KL Div: 3223.6660
Epoch[11/15], Step [410/469], Reconst Loss: 9774.7012, KL Div: 3183.2505
Epoch[11/15], Step [420/469], Reconst Loss: 10321.9609, KL Div: 3184.8350
Epoch[11/15], Step [430/469], Reconst Loss: 10422.4727, KL Div: 3186.4854
Epoch[11/15], Step [440/469], Reconst Loss: 10658.8877, KL Div: 3203.9243
Epoch[11/15], Step [450/469], Reconst Loss: 10263.1055, KL Div: 3264.3481
Epoch[11/15], Step [460/469], Reconst Loss: 10241.3320, KL Div: 3161.2917
Epoch[12/15], Step [10/469], Reconst Loss: 10100.4707, KL Div: 3252.5620
Epoch[12/15], Step [20/469], Reconst Loss: 10362.4365, KL Div: 3182.6909
Epoch[12/15], Step [30/469], Reconst Loss: 10143.4043, KL Div: 3187.7324
Epoch[12/15], Step [40/469], Reconst Loss: 10882.9590, KL Div: 3299.7056
Epoch[12/15], Step [50/469], Reconst Loss: 10397.6016, KL Div: 3282.6797
Epoch[12/15], Step [60/469], Reconst Loss: 10318.3145, KL Div: 3248.0708
Epoch[12/15], Step [70/469], Reconst Loss: 9727.1230, KL Div: 3104.6587
Epoch[12/15], Step [80/469], Reconst Loss: 10144.8438, KL Div: 3335.3586
Epoch[12/15], Step [90/469], Reconst Loss: 10765.9844, KL Div: 3157.8489
Epoch[12/15], Step [100/469], Reconst Loss: 10360.7236, KL Div: 3243.4824
Epoch[12/15], Step [110/469], Reconst Loss: 10216.3037, KL Div: 3199.0112
Epoch[12/15], Step [120/469], Reconst Loss: 10396.5215, KL Div: 3330.6196
Epoch[12/15], Step [130/469], Reconst Loss: 10355.6631, KL Div: 3222.9419
Epoch[12/15], Step [140/469], Reconst Loss: 9937.6602, KL Div: 3146.4893
Epoch[12/15], Step [150/469], Reconst Loss: 9942.9316, KL Div: 3222.0889
Epoch[12/15], Step [160/469], Reconst Loss: 10471.5215, KL Div: 3199.5005
Epoch[12/15], Step [170/469], Reconst Loss: 10320.1836, KL Div: 3271.3301
Epoch[12/15], Step [180/469], Reconst Loss: 9788.8047, KL Div: 3104.6367
Epoch[12/15], Step [190/469], Reconst Loss: 11003.0684, KL Div: 3323.4395
Epoch[12/15], Step [200/469], Reconst Loss: 10676.1084, KL Div: 3284.1753
Epoch[12/15], Step [210/469], Reconst Loss: 10107.3662, KL Div: 3209.1809
Epoch[12/15], Step [220/469], Reconst Loss: 10156.5889, KL Div: 3244.6765
Epoch[12/15], Step [230/469], Reconst Loss: 10257.4248, KL Div: 3279.8970
Epoch[12/15], Step [240/469], Reconst Loss: 10590.3086, KL Div: 3294.7998
Epoch[12/15], Step [250/469], Reconst Loss: 10258.4570, KL Div: 3237.8452
Epoch[12/15], Step [260/469], Reconst Loss: 10488.2148, KL Div: 3423.0193
Epoch[12/15], Step [270/469], Reconst Loss: 10744.1660, KL Div: 3183.3455
Epoch[12/15], Step [280/469], Reconst Loss: 10914.4561, KL Div: 3288.6992
Epoch[12/15], Step [290/469], Reconst Loss: 9729.5820, KL Div: 3215.9641
Epoch[12/15], Step [300/469], Reconst Loss: 10665.6982, KL Div: 3264.9512
Epoch[12/15], Step [310/469], Reconst Loss: 9803.0322, KL Div: 3138.8916
Epoch[12/15], Step [320/469], Reconst Loss: 9962.8525, KL Div: 3285.4980
Epoch[12/15], Step [330/469], Reconst Loss: 10311.7148, KL Div: 3220.7627
Epoch[12/15], Step [340/469], Reconst Loss: 10059.0391, KL Div: 3220.6550
Epoch[12/15], Step [350/469], Reconst Loss: 10087.7500, KL Div: 3218.1709
Epoch[12/15], Step [360/469], Reconst Loss: 10213.5684, KL Div: 3224.1826
Epoch[12/15], Step [370/469], Reconst Loss: 9900.9277, KL Div: 3133.0320
Epoch[12/15], Step [380/469], Reconst Loss: 10240.7793, KL Div: 3247.9814
Epoch[12/15], Step [390/469], Reconst Loss: 10284.7070, KL Div: 3239.2866
Epoch[12/15], Step [400/469], Reconst Loss: 10419.7158, KL Div: 3317.0000
Epoch[12/15], Step [410/469], Reconst Loss: 10537.9961, KL Div: 3295.1472
Epoch[12/15], Step [420/469], Reconst Loss: 10414.0420, KL Div: 3268.8992
Epoch[12/15], Step [430/469], Reconst Loss: 10457.7754, KL Div: 3189.6006
Epoch[12/15], Step [440/469], Reconst Loss: 10238.2861, KL Div: 3341.2368
Epoch[12/15], Step [450/469], Reconst Loss: 10144.2783, KL Div: 3146.1230
Epoch[12/15], Step [460/469], Reconst Loss: 9952.5000, KL Div: 3254.5134
Epoch[13/15], Step [10/469], Reconst Loss: 10561.1523, KL Div: 3278.7744
Epoch[13/15], Step [20/469], Reconst Loss: 10404.3721, KL Div: 3176.3213
Epoch[13/15], Step [30/469], Reconst Loss: 10624.2871, KL Div: 3327.5386
Epoch[13/15], Step [40/469], Reconst Loss: 10325.8320, KL Div: 3344.9980
Epoch[13/15], Step [50/469], Reconst Loss: 10723.1426, KL Div: 3287.0120
Epoch[13/15], Step [60/469], Reconst Loss: 10437.0645, KL Div: 3357.3662
Epoch[13/15], Step [70/469], Reconst Loss: 10004.9355, KL Div: 3170.4160
Epoch[13/15], Step [80/469], Reconst Loss: 10082.7598, KL Div: 3213.1504
Epoch[13/15], Step [90/469], Reconst Loss: 10512.5586, KL Div: 3321.3447
Epoch[13/15], Step [100/469], Reconst Loss: 10059.7617, KL Div: 3234.3960
Epoch[13/15], Step [110/469], Reconst Loss: 10556.2148, KL Div: 3303.7673
Epoch[13/15], Step [120/469], Reconst Loss: 10303.7070, KL Div: 3234.6082
Epoch[13/15], Step [130/469], Reconst Loss: 10065.9121, KL Div: 3103.7578
Epoch[13/15], Step [140/469], Reconst Loss: 10536.8574, KL Div: 3366.4834
Epoch[13/15], Step [150/469], Reconst Loss: 10153.4932, KL Div: 3125.0923
Epoch[13/15], Step [160/469], Reconst Loss: 9987.3643, KL Div: 3284.4302
Epoch[13/15], Step [170/469], Reconst Loss: 10263.2109, KL Div: 3213.4697
Epoch[13/15], Step [180/469], Reconst Loss: 10746.4902, KL Div: 3392.5688
Epoch[13/15], Step [190/469], Reconst Loss: 10265.0459, KL Div: 3304.5281
Epoch[13/15], Step [200/469], Reconst Loss: 10162.6172, KL Div: 3181.8455
Epoch[13/15], Step [210/469], Reconst Loss: 10058.0820, KL Div: 3228.5020
Epoch[13/15], Step [220/469], Reconst Loss: 10106.4180, KL Div: 3194.3140
Epoch[13/15], Step [230/469], Reconst Loss: 9975.6738, KL Div: 3194.1494
Epoch[13/15], Step [240/469], Reconst Loss: 10023.4424, KL Div: 3229.2405
Epoch[13/15], Step [250/469], Reconst Loss: 10043.7246, KL Div: 3207.5303
Epoch[13/15], Step [260/469], Reconst Loss: 9962.7070, KL Div: 3155.2725
Epoch[13/15], Step [270/469], Reconst Loss: 10077.5498, KL Div: 3176.8755
Epoch[13/15], Step [280/469], Reconst Loss: 10064.2910, KL Div: 3234.5818
Epoch[13/15], Step [290/469], Reconst Loss: 10216.9326, KL Div: 3207.9661
Epoch[13/15], Step [300/469], Reconst Loss: 10207.9873, KL Div: 3253.5781
Epoch[13/15], Step [310/469], Reconst Loss: 10112.1582, KL Div: 3152.4810
Epoch[13/15], Step [320/469], Reconst Loss: 10220.6309, KL Div: 3243.7036
Epoch[13/15], Step [330/469], Reconst Loss: 10113.3047, KL Div: 3277.7246
Epoch[13/15], Step [340/469], Reconst Loss: 10360.3691, KL Div: 3235.7180
Epoch[13/15], Step [350/469], Reconst Loss: 10314.1572, KL Div: 3298.3831
Epoch[13/15], Step [360/469], Reconst Loss: 9738.2354, KL Div: 3156.9714
Epoch[13/15], Step [370/469], Reconst Loss: 10193.6719, KL Div: 3227.9404
Epoch[13/15], Step [380/469], Reconst Loss: 10099.7305, KL Div: 3256.4912
Epoch[13/15], Step [390/469], Reconst Loss: 9898.5684, KL Div: 3212.1011
Epoch[13/15], Step [400/469], Reconst Loss: 10243.5762, KL Div: 3353.5847
Epoch[13/15], Step [410/469], Reconst Loss: 10356.5869, KL Div: 3183.2246
Epoch[13/15], Step [420/469], Reconst Loss: 10546.1289, KL Div: 3293.1250
Epoch[13/15], Step [430/469], Reconst Loss: 9828.0391, KL Div: 3263.3047
Epoch[13/15], Step [440/469], Reconst Loss: 9905.3340, KL Div: 3148.6338
Epoch[13/15], Step [450/469], Reconst Loss: 10301.7832, KL Div: 3251.5398
Epoch[13/15], Step [460/469], Reconst Loss: 10545.7549, KL Div: 3337.8765
Epoch[14/15], Step [10/469], Reconst Loss: 10069.1475, KL Div: 3265.8403
Epoch[14/15], Step [20/469], Reconst Loss: 10634.7285, KL Div: 3417.9934
Epoch[14/15], Step [30/469], Reconst Loss: 10381.9365, KL Div: 3184.7178
Epoch[14/15], Step [40/469], Reconst Loss: 10487.7070, KL Div: 3267.9077
Epoch[14/15], Step [50/469], Reconst Loss: 10231.0342, KL Div: 3363.9539
Epoch[14/15], Step [60/469], Reconst Loss: 10475.6465, KL Div: 3279.8052
Epoch[14/15], Step [70/469], Reconst Loss: 10026.4590, KL Div: 3185.8557
Epoch[14/15], Step [80/469], Reconst Loss: 10289.2051, KL Div: 3390.5732
Epoch[14/15], Step [90/469], Reconst Loss: 10010.5078, KL Div: 3207.6255
Epoch[14/15], Step [100/469], Reconst Loss: 10236.8457, KL Div: 3321.0464
Epoch[14/15], Step [110/469], Reconst Loss: 10333.8789, KL Div: 3246.1592
Epoch[14/15], Step [120/469], Reconst Loss: 10373.4316, KL Div: 3179.2629
Epoch[14/15], Step [130/469], Reconst Loss: 9921.7148, KL Div: 3174.9546
Epoch[14/15], Step [140/469], Reconst Loss: 10403.0566, KL Div: 3209.5942
Epoch[14/15], Step [150/469], Reconst Loss: 10434.0850, KL Div: 3237.9197
Epoch[14/15], Step [160/469], Reconst Loss: 9863.9453, KL Div: 3282.2021
Epoch[14/15], Step [170/469], Reconst Loss: 10278.5703, KL Div: 3328.7625
Epoch[14/15], Step [180/469], Reconst Loss: 9605.9189, KL Div: 3219.9016
Epoch[14/15], Step [190/469], Reconst Loss: 9991.4883, KL Div: 3135.2844
Epoch[14/15], Step [200/469], Reconst Loss: 9956.5195, KL Div: 3196.9170
Epoch[14/15], Step [210/469], Reconst Loss: 10533.9990, KL Div: 3343.3950
Epoch[14/15], Step [220/469], Reconst Loss: 10042.6260, KL Div: 3177.3831
Epoch[14/15], Step [230/469], Reconst Loss: 9880.4521, KL Div: 3222.9128
Epoch[14/15], Step [240/469], Reconst Loss: 10170.3877, KL Div: 3273.3689
Epoch[14/15], Step [250/469], Reconst Loss: 10437.9170, KL Div: 3242.1714
Epoch[14/15], Step [260/469], Reconst Loss: 10217.8145, KL Div: 3159.4246
Epoch[14/15], Step [270/469], Reconst Loss: 9837.6758, KL Div: 3148.0342
Epoch[14/15], Step [280/469], Reconst Loss: 10033.9102, KL Div: 3285.6162
Epoch[14/15], Step [290/469], Reconst Loss: 10050.0527, KL Div: 3267.7573
Epoch[14/15], Step [300/469], Reconst Loss: 10737.7852, KL Div: 3276.8811
Epoch[14/15], Step [310/469], Reconst Loss: 10654.3281, KL Div: 3195.3047
Epoch[14/15], Step [320/469], Reconst Loss: 9751.8867, KL Div: 3234.4192
Epoch[14/15], Step [330/469], Reconst Loss: 9862.9922, KL Div: 3285.5381
Epoch[14/15], Step [340/469], Reconst Loss: 9958.1113, KL Div: 3158.3511
Epoch[14/15], Step [350/469], Reconst Loss: 10145.1914, KL Div: 3199.6873
Epoch[14/15], Step [360/469], Reconst Loss: 10368.6338, KL Div: 3395.5317
Epoch[14/15], Step [370/469], Reconst Loss: 10013.0508, KL Div: 3153.7812
Epoch[14/15], Step [380/469], Reconst Loss: 10253.8887, KL Div: 3274.3643
Epoch[14/15], Step [390/469], Reconst Loss: 10283.4805, KL Div: 3308.3999
Epoch[14/15], Step [400/469], Reconst Loss: 10379.4180, KL Div: 3318.8638
Epoch[14/15], Step [410/469], Reconst Loss: 9914.1719, KL Div: 3175.6028
Epoch[14/15], Step [420/469], Reconst Loss: 10168.4199, KL Div: 3187.2297
Epoch[14/15], Step [430/469], Reconst Loss: 10245.3184, KL Div: 3236.4622
Epoch[14/15], Step [440/469], Reconst Loss: 10100.7812, KL Div: 3267.4075
Epoch[14/15], Step [450/469], Reconst Loss: 9914.7559, KL Div: 3134.0366
Epoch[14/15], Step [460/469], Reconst Loss: 10017.5098, KL Div: 3430.6792
Epoch[15/15], Step [10/469], Reconst Loss: 10327.4453, KL Div: 3295.9519
Epoch[15/15], Step [20/469], Reconst Loss: 10625.3418, KL Div: 3216.5332
Epoch[15/15], Step [30/469], Reconst Loss: 10300.7256, KL Div: 3204.3267
Epoch[15/15], Step [40/469], Reconst Loss: 10053.3760, KL Div: 3213.3494
Epoch[15/15], Step [50/469], Reconst Loss: 9802.7109, KL Div: 3277.8635
Epoch[15/15], Step [60/469], Reconst Loss: 10406.4102, KL Div: 3174.1592
Epoch[15/15], Step [70/469], Reconst Loss: 10141.3604, KL Div: 3269.6350
Epoch[15/15], Step [80/469], Reconst Loss: 10166.9277, KL Div: 3256.5972
Epoch[15/15], Step [90/469], Reconst Loss: 10143.2266, KL Div: 3141.7861
Epoch[15/15], Step [100/469], Reconst Loss: 10453.3926, KL Div: 3279.7358
Epoch[15/15], Step [110/469], Reconst Loss: 10315.9658, KL Div: 3261.2327
Epoch[15/15], Step [120/469], Reconst Loss: 9827.2764, KL Div: 3202.5688
Epoch[15/15], Step [130/469], Reconst Loss: 10371.4766, KL Div: 3302.3269
Epoch[15/15], Step [140/469], Reconst Loss: 10176.9668, KL Div: 3277.7471
Epoch[15/15], Step [150/469], Reconst Loss: 10535.4043, KL Div: 3319.8499
Epoch[15/15], Step [160/469], Reconst Loss: 10384.1865, KL Div: 3258.4556
Epoch[15/15], Step [170/469], Reconst Loss: 10176.8857, KL Div: 3313.0811
Epoch[15/15], Step [180/469], Reconst Loss: 9736.1553, KL Div: 3179.9526
Epoch[15/15], Step [190/469], Reconst Loss: 10388.7783, KL Div: 3210.8306
Epoch[15/15], Step [200/469], Reconst Loss: 10706.0146, KL Div: 3314.2937
Epoch[15/15], Step [210/469], Reconst Loss: 10436.8965, KL Div: 3201.7385
Epoch[15/15], Step [220/469], Reconst Loss: 10383.0273, KL Div: 3267.3047
Epoch[15/15], Step [230/469], Reconst Loss: 10396.6680, KL Div: 3188.7212
Epoch[15/15], Step [240/469], Reconst Loss: 10093.4375, KL Div: 3226.1409
Epoch[15/15], Step [250/469], Reconst Loss: 9976.1230, KL Div: 3331.6909
Epoch[15/15], Step [260/469], Reconst Loss: 9936.5430, KL Div: 3166.5913
Epoch[15/15], Step [270/469], Reconst Loss: 10188.4463, KL Div: 3264.0513
Epoch[15/15], Step [280/469], Reconst Loss: 9759.2363, KL Div: 3212.6204
Epoch[15/15], Step [290/469], Reconst Loss: 10251.6484, KL Div: 3285.7480
Epoch[15/15], Step [300/469], Reconst Loss: 10107.9277, KL Div: 3225.7773
Epoch[15/15], Step [310/469], Reconst Loss: 10231.5732, KL Div: 3285.9448
Epoch[15/15], Step [320/469], Reconst Loss: 10227.3945, KL Div: 3287.7163
Epoch[15/15], Step [330/469], Reconst Loss: 10169.8535, KL Div: 3291.9043
Epoch[15/15], Step [340/469], Reconst Loss: 10114.7881, KL Div: 3198.2019
Epoch[15/15], Step [350/469], Reconst Loss: 9916.6758, KL Div: 3298.3027
Epoch[15/15], Step [360/469], Reconst Loss: 10149.1543, KL Div: 3334.2546
Epoch[15/15], Step [370/469], Reconst Loss: 10391.2930, KL Div: 3154.2920
Epoch[15/15], Step [380/469], Reconst Loss: 10285.0098, KL Div: 3314.5752
Epoch[15/15], Step [390/469], Reconst Loss: 10037.2598, KL Div: 3348.3552
Epoch[15/15], Step [400/469], Reconst Loss: 10347.6201, KL Div: 3230.4548
Epoch[15/15], Step [410/469], Reconst Loss: 10050.3828, KL Div: 3257.5193
Epoch[15/15], Step [420/469], Reconst Loss: 10073.8594, KL Div: 3217.1846
Epoch[15/15], Step [430/469], Reconst Loss: 9558.9961, KL Div: 3181.1758
Epoch[15/15], Step [440/469], Reconst Loss: 9885.7969, KL Div: 3207.2212
Epoch[15/15], Step [450/469], Reconst Loss: 10039.3613, KL Div: 3216.7864
Epoch[15/15], Step [460/469], Reconst Loss: 10072.1973, KL Div: 3270.6738

在这里插入图片描述
这里分别是采样数据和重建数据实例。

在这里插入图片描述

# save and load the model checkpoint.
torch.save(model.state_dict(), 'model_param.ckpt')
model.load_state_dict(torch.load('model_param.ckpt'))

<All keys matched successfully>
AIHUBEI
关注
  • 0
    点赞
  • 踩
  • 1
    收藏
    觉得还不错? 一键收藏
  • 0
    评论
专栏目录
PyTorch深度学习实战（28）——卷积自编码器(Convolutional Autoencoder)
盼小辉丶的博客
12-21 3634
卷积自编码器是一种基于卷积神经网络结构的自编码器，适用于处理图像数据。在本节中，我们介绍了卷积自编码器的模型架构，使用 PyTorch 从零开始实现在 MNIST 数据集上训练了一个简单的卷积自编码器，并使用 t-SNE 技术在二维平面中表示了所有测试图像的潜变量。
pytorch实现变分自编码器
12-20
pytorch实现变分自编码器
变分自编码器——pytorch实现VAE
qq_36892712的博客
08-04 1827
变分自编码器（Variational Autoencoder，VAE）是一种生成模型，结合了自编码器和变分推断的思想。它可以用于学习数据的潜在表示，并用于生成新的样本。VAE的目标是学习一个潜在空间（latent space），使得原始数据可以通过从该潜在空间中采样并解码重构出来。编码器（encoder）和解码器（decoder）。编码器将输入数据映射到潜在空间中的一组潜在变量（latent variables）。编码器将输入数据xxx映射到潜在变量zzz的概率分布qz∣x。
pytorch-deeplab V3+训练自己的数据集 问题总结
gsgs1234的博客
03-27 3612
pytorch+deeplab V3训练自己的数据集 问题总结 进行植被分类实验，使用pytorch版本的deeplabV3训练自己的数据集，训练过程参考大大的博客： https://blog.csdn.net/qq_39056987/article/details/106455828 一、安装包 修改了各种配置之后，先根据标红安装各种缺少的包，代码中无明显错误后开始训练。 二、训练 （1）RuntimeError: unexpected EOF, expected 24460 more bytes. Th
Pytorch搭建基于LSTM的Variational AutoEncoder(变分自编码器)
weixin_43895314的博客
08-23 580
【代码】【无标题】
Pytorch实现变分自编码器（VAE）
阿伟的博客
09-23 2105
变分自编码器是自编码器的变种，是一种生成模型。
【Pytorch深度学习实战】（11）变分自动编码器（VAE）
热门推荐
sikh_0529的博客
09-20 1万+
而变分自编码器便是用“取值的概率分布”代替原先的单值来描述对特征的观察的模型，如下图的右边部分所示，经过变分自编码器的编码，每张图片的微笑特征不再是自编码器中的单值而是一个概率分布。另一个根据生成的隐变量变分概率分布，还原生成原始数据的近似概率分布，称为生成网络。在上面的模型中，经过反复训练，我们的输入数据X最终被转化为一个编码向量X’， 其中X’的每个维度表示一些学到的关于数据的特征，而X’在每个维度上的取值代表X在该特征上的表现。通过上述的编解码过程，我们实质上实施了连续，平滑的潜在空间表示。
Pytorch从0到1之变分自编码器——(10)
weixin_44755413的博客
05-04 419
开篇 这次我们来说一说变分自编码器。变分编码器也是一种很常见的网络结构。它的作用和GAN有些类似，都是为我们生成一张可以"以假乱真"的图片。但是VAE与GAN不同的是，它不用区分生成器和区分器，他在一个网络中完成整个过程。 我们首先输入图片，对他进编码，然后通过我们的网络结构生成编码的方差与均值，然后再解码生成图片，这里最重要的是这个方差和均值的生成。自己刚刚复现了一遍，感觉这里还是挺多需要了解和...
变分自编码器VAE实现MNIST数据集生成by Pytorch
winycg的博客
05-18 1万+
Recap自编码器： 自编码器中，需要输入一个原始图片，原始图片经过编码之后得到一个隐向量，隐向量解码产生原图片对应的图片。在这种情况下，只能生成原图片对应的图片而无法任意生成新的图片，因为隐向量都是原始图片确定的。 变分自编码器VAE 引入变分自编码器（Variational autoencoder）可以在遵循某一分布下随机产生一些隐向量来生成与原始图片不相同的图片，而不需要预先给定原始图片。为...
Pytorch 如何 优化/调整 模型参数
xiaoxifei的专栏
04-30 9595
Pytorch 如何自动优化/调整 模型超参 文章目录Pytorch 如何自动优化/调整 模型超参背景优化模型参数贝叶斯优化深度学习框架下的参数优化平台安装使用参考参考 背景 对于优化模型性能表现而言，主要可归纳为两种方式： 采用NAS的方式搜索，获得某个任务上最优的模型结构以及模型参数设置 优化模型参数 诚然，对于第一种方式来说，NAS对算力的消耗是巨大的，因而需要具备巨量计算资源才能够进行...
vae.pytorch：深度特征一致的变分自动编码器的PyTorch实现
02-22
PyTorch中的深度特征一致变分自动编码器 深度特征一致变分自动编码器的PyTorch实现。 我基于侯宪旭，沉琳琳，孙可，邱国平实施了DFC-VAE。 我使用CelebA数据集训练了该模型。 有关数据集的更多详细信息，请。 安装 克隆此存储库。 建议使用python 3.6。 使用命令pip install -r requirements.txt安装库。 数据集 您需要从上下载CelebA数据集，并按照以下方式进行排列。 . ├── README.md ├── requirements.txt ├── models ├── utils ... └── data └── celeba ├── images │ ├── 000001.jpg │ └── ... └── annotations
Pytorch实现VAE变分自动编码器生成MNIST手写数字图像
06-11
# Pytorch实现VAE变分自动编码器生成MNIST手写数字图像 1. VAE模型的Pytorch源码，训练后其解码器就是生成模型； 2. 在MNIST数据集上训练了50个epochs，训练过程的生成效果放在result文件夹下，训练后的模型保存为...
Pytorch变分自动编码器vae代码.zip
08-13
Pytorch变分自动编码器vae代码.
vqvae:矢量量化变分自动编码器（https的pytorch实现
05-04
矢量量化变分自编码器这是矢量量化变分自动编码器（ ）的PyTorch实现。 您可以找到作者的带有。安装依赖项要安装依赖项，请使用Python 3创建一个conda或虚拟环境，然后运行pip install -r requirements.txt 。运行VQ...
人工智能入门（一）：基于Pytorch的手写数字识别模型
weixin_63441326的博客
04-21 800
学习过线性代数的话，你可能对矩阵比较熟悉，张量其实就是矩阵向任意维度的推广。（1）如果一个矩阵只有一个元素，那么它被称为0阶张量；（2）如果一个矩阵只有一行或者一列（我们称其为只有一个“轴”），则被称为一阶张量；（3）显然，大部分的矩阵都有m行n列（有两个轴），则就被称为二阶张量；（4）如果将一个二阶张量看作一个整体，由多个二阶张量组成一个新的数组，则二阶张量就被推广为三阶张量；当然，我本人有个小技巧帮助你推导三阶之后的张量。来，跟我念：一个张量、往下拉、往右拉、往后拉；
6.Vgg16--CNN经典网络模型详解（pytorch实现）
weixin_71719718的博客
04-24 352
3.下载完后写一个spile_data.py文件，将数据集进行分类。4.再写一个train.py文件，用来训练模型。
PyTorch深度解析：Tensor——神经网络的核心构建块
huanfeng_AI的博客
04-20 1116
具体来说，神经网络的输入数据被转换为Tensor，然后通过一系列的网络层进行前向传播，得到输出Tensor。在训练过程中，根据输出Tensor与真实标签之间的误差，通过反向传播算法计算梯度，并更新网络参数（这些参数也是以Tensor的形式存在的）。本文将深入探讨PyTorch中的Tensor，从其基本定义、特性、操作到实际应用，全面解析其在神经网络中的核心地位。：PyTorch中的Tensor支持多种数据类型，如float32、float64、int32等，以适应不同的计算需求。
分析和比较深度学习框架 PyTorch 和 Tensorflow
weixin_43784341的博客
04-24 582
综上所述，PyTorch和TensorFlow都是强大的深度学习框架，具有各自的优点和适用场景。选择哪个框架取决于你的具体需求、经验以及对框架的偏好。如果你正在寻找一个易于使用和调试的框架，并且更关注实验和原型设计，那么PyTorch可能是一个更好的选择。而如果你需要高性能的分布式训练、强大的硬件支持和丰富的生态系统，那么TensorFlow可能更适合你。PyTorch和TensorFlow都是目前最受欢迎的深度学习框架，它们各自具有独特的优点和特性。
【炼金术士】BatchSize对网络训练的影响
最新发布
kabuto_hui的博客
04-25 1047
首先我们设置一个非常小的初始学习率，比如1e-5，然后在每个batch之后都更新网络，同时增加学习率，统计每个batch计算出的loss。最后我们可以描绘出学习的变化曲线和loss的变化曲线，从中就能够发现最好的学习率。针对BatchSize越大，网络训练速度快这条很好理解，在相同的单位时间内，一次训练的数据越多，网络训练的速度也就越快。初始的学习率肯定是有一个最优值的，过大则导致模型不收敛，过小则导致模型收敛特别慢或者无法学习，下图展示了不同大小的学习率下模型收敛情况的可能性。
pytorch实现变分自动编码器
12-06
以下是使用PyTorch实现变分自动编码器的步骤： 1.导入所需的库和模块，包括torch、torch.nn、torch.nn.functional、torchvision等。 2.定义设备配置，判断是否有可用的GPU，如果有则使用GPU，否则使用CPU。 3.定义变分自动编码器的编码器和解码器。编码器由两个全连接层和一个输出层组成，解码器由一个全连接层和一个输出层组成。 4.定义变分自动编码器的前向传播函数forward()，其中包括编码器和解码器的前向传播过程。 5.定义变分自动编码器的损失函数，包括重构误差和KL散度。 6.定义优化器，使用Adam优化器。 7.训练模型，包括前向传播、计算损失、反向传播和优化器更新参数。 8.保存模型和生成样本图片。 下面是完整的代码实现： ```python import os import torch import torch.nn as nn import torch.nn.functional as F import torchvision from torchvision import transforms from torchvision.utils import save_image import matplotlib.pyplot as plt # 设备配置 device = torch.device('cuda' if torch.cuda.is_available() else 'cpu') # 定义变分自动编码器的编码器和解码器 class VAE(nn.Module): def __init__(self): super(VAE, self).__init__() self.fc1 = nn.Linear(784, 400) self.fc21 = nn.Linear(400, 20) self.fc22 = nn.Linear(400, 20) self.fc3 = nn.Linear(20, 400) self.fc4 = nn.Linear(400, 784) def encode(self, x): h1 = F.relu(self.fc1(x)) return self.fc21(h1), self.fc22(h1) def reparameterize(self, mu, logvar): std = torch.exp(0.5*logvar) eps = torch.randn_like(std) return mu + eps*std def decode(self, z): h3 = F.relu(self.fc3(z)) return torch.sigmoid(self.fc4(h3)) def forward(self, x): mu, logvar = self.encode(x.view(-1, 784)) z = self.reparameterize(mu, logvar) return self.decode(z), mu, logvar # 定义变分自动编码器的损失函数 def loss_function(recon_x, x, mu, logvar): BCE = F.binary_cross_entropy(recon_x, x.view(-1, 784), reduction='sum') KLD = -0.5 * torch.sum(1 + logvar - mu.pow(2) - logvar.exp()) return BCE + KLD # 定义优化器 vae = VAE().to(device) optimizer = torch.optim.Adam(vae.parameters(), lr=1e-3) # 训练模型 def train(epoch): vae.train() train_loss = 0 for batch_idx, (data, _) in enumerate(train_loader): data = data.to(device) optimizer.zero_grad() recon_batch, mu, logvar = vae(data) loss = loss_function(recon_batch, data, mu, logvar) loss.backward() train_loss += loss.item() optimizer.step() if batch_idx % 100 == 0: print('Train Epoch: {} [{}/{} ({:.0f}%)]\tLoss: {:.6f}'.format( epoch, batch_idx * len(data), len(train_loader.dataset), 100. * batch_idx / len(train_loader), loss.item() / len(data))) print('====> Epoch: {} Average loss: {:.4f}'.format( epoch, train_loss / len(train_loader.dataset))) # 保存模型和生成样本图片 if not os.path.exists('./vae_samples'): os.mkdir('./vae_samples') def save_samples(epoch): with torch.no_grad(): sample = torch.randn(64, 20).to(device) sample = vae.decode(sample).cpu() save_image(sample.view(64, 1, 28, 28), './vae_samples/sample_' + str(epoch) + '.png') # 加载MNIST数据集 batch_size = 128 transform = transforms.Compose([ transforms.ToTensor(), transforms.Normalize((0.1307,), (0.3081,)) ]) train_dataset = torchvision.datasets.MNIST(root='./data', train=True, transform=transform, download=True) train_loader = torch.utils.data.DataLoader(train_dataset, batch_size=batch_size, shuffle=True) # 训练模型并保存样本图片 for epoch in range(1, 21): train(epoch) save_samples(epoch) ```

“相关推荐”对你有帮助么？

  • 非常没帮助
  • 没帮助
  • 一般
  • 有帮助
  • 非常有帮助
提交
评论
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付：10.00元
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明：

1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载，可以购买VIP、付费专栏及课程。

余额充值