2024-02-08 19:09:18

OpenDocCN · Feb 8, 2024 · 702167c · 702167c
1 parent cca2c40
commit 702167c
Showing 1 changed file with 41 additions and 0 deletions.
diff --git a/totrans/gen-dl_11.yaml b/totrans/gen-dl_11.yaml
@@ -1391,31 +1391,37 @@
   id: totrans-163
   prefs: []
   type: TYPE_NORMAL
+  zh: 这个解决方案最初在2015年He等人的ResNet论文中首次提出。通过在主要加权层周围包含一个跳跃连接*高速公路*，块可以选择绕过复杂的权重更新，简单地通过恒等映射传递。这使得网络可以在不牺牲梯度大小或网络准确性的情况下进行深度训练。
 - en: A diagram of a `ResidualBlock` is shown in [Figure 8-10](#diffusion_residual).
     Note that in some residual blocks, we also include an extra `Conv2D` layer with
     kernel size 1 on the skip connection, to bring the number of channels in line
     with the rest of the block.
   id: totrans-164
   prefs: []
   type: TYPE_NORMAL
+  zh: '`ResidualBlock`的图示如[图8-10](#diffusion_residual)所示。请注意，在一些残差块中，我们还在跳跃连接上包含一个额外的具有核大小1的`Conv2D`层，以使通道数与块的其余部分保持一致。'
 - en: '![](Images/gdl2_0810.png)'
   id: totrans-165
   prefs: []
   type: TYPE_IMG
+  zh: '![](Images/gdl2_0810.png)'
 - en: Figure 8-10\. The `ResidualBlock` in the U-Net
   id: totrans-166
   prefs:
   - PREF_H6
   type: TYPE_NORMAL
+  zh: 图8-10。U-Net中的`ResidualBlock`
 - en: We can code a `ResidualBlock` in Keras as shown in [Example 8-8](#diffusion_residual_code).
   id: totrans-167
   prefs: []
   type: TYPE_NORMAL
+  zh: 我们可以像[示例8-8](#diffusion_residual_code)中所示的那样在Keras中编写`ResidualBlock`。
 - en: Example 8-8\. Code for the `ResidualBlock` in the U-Net
   id: totrans-168
   prefs:
   - PREF_H5
   type: TYPE_NORMAL
+  zh: 示例8-8。U-Net中的`ResidualBlock`代码
 - en: '[PRE7]'
   id: totrans-169
   prefs: []
@@ -1425,50 +1431,60 @@
   id: totrans-170
   prefs: []
   type: TYPE_NORMAL
+  zh: '[![1](Images/1.png)](#co_diffusion_models_CO6-1)'
 - en: Check if the number of channels in the input matches the number of channels
     that we would like the block to output. If not, include an extra `Conv2D` layer
     on the skip connection to bring the number of channels in line with the rest of
     the block.
   id: totrans-171
   prefs: []
   type: TYPE_NORMAL
+  zh: 检查输入中的通道数是否与我们希望该块输出的通道数匹配。如果不匹配，可以在跳跃连接上包含额外的`Conv2D`层，以使通道数与块的其余部分保持一致。
 - en: '[![2](Images/2.png)](#co_diffusion_models_CO6-2)'
   id: totrans-172
   prefs: []
   type: TYPE_NORMAL
+  zh: '[![2](Images/2.png)](#co_diffusion_models_CO6-2)'
 - en: Apply a `BatchNormalization` layer.
   id: totrans-173
   prefs: []
   type: TYPE_NORMAL
+  zh: 应用`BatchNormalization`层。
 - en: '[![3](Images/3.png)](#co_diffusion_models_CO6-3)'
   id: totrans-174
   prefs: []
   type: TYPE_NORMAL
+  zh: '[![3](Images/3.png)](#co_diffusion_models_CO6-3)'
 - en: Apply two `Conv2D` layers.
   id: totrans-175
   prefs: []
   type: TYPE_NORMAL
+  zh: 应用两个`Conv2D`层。
 - en: '[![4](Images/4.png)](#co_diffusion_models_CO6-4)'
   id: totrans-176
   prefs: []
   type: TYPE_NORMAL
+  zh: '[![4](Images/4.png)](#co_diffusion_models_CO6-4)'
 - en: Add the original block input to the output to provide the final output from
     the block.
   id: totrans-177
   prefs: []
   type: TYPE_NORMAL
+  zh: 将原始块输入添加到输出中，以提供块的最终输出。
 - en: DownBlocks and UpBlocks
   id: totrans-178
   prefs:
   - PREF_H3
   type: TYPE_NORMAL
+  zh: DownBlocks和UpBlocks
 - en: Each successive `DownBlock` increases the number of channels via `block_depth`
     (=2 in our example) `ResidualBlock`s, while also applying a final `AveragePooling2D`
     layer in order to halve the size of the image. Each `ResidualBlock` is added to
     a list for use later by the `UpBlock` layers as skip connections across the U-Net.
   id: totrans-179
   prefs: []
   type: TYPE_NORMAL
+  zh: 每个连续的`DownBlock`通过`block_depth`（在我们的示例中为2）`ResidualBlock`增加通道数，同时还应用最终的`AveragePooling2D`层以将图像尺寸减半。每个`ResidualBlock`都添加到一个列表中，以便稍后由`UpBlock`层作为U-Net中的跳跃连接使用。
 - en: An `UpBlock` first applies an `UpSampling2D` layer that doubles the size of
     the image, through bilinear interpolation. Each successive `UpBlock` decreases
     the number of channels via `block_depth` (=2) `ResidualBlock`s, while also concatenating
@@ -1477,24 +1493,29 @@
   id: totrans-180
   prefs: []
   type: TYPE_NORMAL
+  zh: '`UpBlock`首先应用一个`UpSampling2D`层，通过双线性插值将图像大小加倍。每个连续的`UpBlock`通过`block_depth`（=2）`ResidualBlock`减少通道数，同时还通过U-Net中的跳跃连接将`DownBlock`的输出连接起来。这个过程的图示如[图8-11](#diffusion_down_up_block)所示。'
 - en: '![](Images/gdl2_0811.png)'
   id: totrans-181
   prefs: []
   type: TYPE_IMG
+  zh: '![](Images/gdl2_0811.png)'
 - en: Figure 8-11\. The `DownBlock` and corresponding `UpBlock` in the U-Net
   id: totrans-182
   prefs:
   - PREF_H6
   type: TYPE_NORMAL
+  zh: 图8-11。U-Net中的`DownBlock`和相应的`UpBlock`
 - en: We can code the `DownBlock` and `UpBlock` using Keras as illustrated in [Example 8-9](#diffusion_down_up_code).
   id: totrans-183
   prefs: []
   type: TYPE_NORMAL
+  zh: 我们可以使用Keras编写`DownBlock`和`UpBlock`，如[示例8-9](#diffusion_down_up_code)所示。
 - en: Example 8-9\. Code for the `DownBlock` and `UpBlock` in the U-Net model
   id: totrans-184
   prefs:
   - PREF_H5
   type: TYPE_NORMAL
+  zh: 示例8-9。U-Net模型中的`DownBlock`和`UpBlock`代码
 - en: '[PRE8]'
   id: totrans-185
   prefs: []
@@ -1504,71 +1525,86 @@
   id: totrans-186
   prefs: []
   type: TYPE_NORMAL
+  zh: '[![1](Images/1.png)](#co_diffusion_models_CO7-1)'
 - en: The `DownBlock` increases the number of channels in the image using a `ResidualBlock`
     of a given `width`…
   id: totrans-187
   prefs: []
   type: TYPE_NORMAL
+  zh: '`DownBlock`通过给定`width`的`ResidualBlock`增加图像中的通道数…'
 - en: '[![2](Images/2.png)](#co_diffusion_models_CO7-2)'
   id: totrans-188
   prefs: []
   type: TYPE_NORMAL
+  zh: '[![2](Images/2.png)](#co_diffusion_models_CO7-2)'
 - en: …each of which are saved to a list (`skips`) for use later by the `UpBlock`s.
   id: totrans-189
   prefs: []
   type: TYPE_NORMAL
+  zh: …每个都保存在一个列表（`skips`）中，以便稍后由`UpBlock`使用。
 - en: '[![3](Images/3.png)](#co_diffusion_models_CO7-3)'
   id: totrans-190
   prefs: []
   type: TYPE_NORMAL
+  zh: '[![3](Images/3.png)](#co_diffusion_models_CO7-3)'
 - en: A final `AveragePooling2D` layer reduces the dimensionality of the image by
     half.
   id: totrans-191
   prefs: []
   type: TYPE_NORMAL
+  zh: 最终的`AveragePooling2D`层将图像的维度减半。
 - en: '[![4](Images/4.png)](#co_diffusion_models_CO7-4)'
   id: totrans-192
   prefs: []
   type: TYPE_NORMAL
+  zh: '[![4](Images/4.png)](#co_diffusion_models_CO7-4)'
 - en: The `UpBlock` begins with an `UpSampling2D` layer that doubles the size of the
     image.
   id: totrans-193
   prefs: []
   type: TYPE_NORMAL
+  zh: '`UpBlock`从一个`UpSampling2D`层开始，将图像大小加倍。'
 - en: '[![5](Images/5.png)](#co_diffusion_models_CO7-5)'
   id: totrans-194
   prefs: []
   type: TYPE_NORMAL
+  zh: '[![5](Images/5.png)](#co_diffusion_models_CO7-5)'
 - en: The output from a `DownBlock` layer is glued to the current output using a `Concatenate`
     layer.
   id: totrans-195
   prefs: []
   type: TYPE_NORMAL
+  zh: '`DownBlock`层的输出通过`Concatenate`层连接到当前输出。'
 - en: '[![6](Images/6.png)](#co_diffusion_models_CO7-6)'
   id: totrans-196
   prefs: []
   type: TYPE_NORMAL
+  zh: '[![6](Images/6.png)](#co_diffusion_models_CO7-6)'
 - en: A `ResidualBlock` is used to reduce the number of channels in the image as it
     passes through the `UpBlock`.
   id: totrans-197
   prefs: []
   type: TYPE_NORMAL
+  zh: '`ResidualBlock`用于在图像通过`UpBlock`时减少通道数。'
 - en: Training the Diffusion Model
   id: totrans-198
   prefs:
   - PREF_H2
   type: TYPE_NORMAL
+  zh: 训练扩散模型
 - en: We now have all the components in place to train our denoising diffusion model!
     [Example 8-10](#diffusion_train_code) creates, compiles, and fits the diffusion
     model.
   id: totrans-199
   prefs: []
   type: TYPE_NORMAL
+  zh: 现在我们已经准备好训练我们的去噪扩散模型了！[示例8-10](#diffusion_train_code)创建、编译和拟合扩散模型。
 - en: Example 8-10\. Code for training the `DiffusionModel`
   id: totrans-200
   prefs:
   - PREF_H5
   type: TYPE_NORMAL
+  zh: 示例8-10。训练`DiffusionModel`的代码
 - en: '[PRE9]'
   id: totrans-201
   prefs: []
@@ -1578,24 +1614,29 @@
   id: totrans-202
   prefs: []
   type: TYPE_NORMAL
+  zh: '[![1](Images/1.png)](#co_diffusion_models_CO8-1)'
 - en: Instantiate the model.
   id: totrans-203
   prefs: []
   type: TYPE_NORMAL
+  zh: 实例化模型。
 - en: '[![2](Images/2.png)](#co_diffusion_models_CO8-2)'
   id: totrans-204
   prefs: []
   type: TYPE_NORMAL
+  zh: '[![2](Images/2.png)](#co_diffusion_models_CO8-2)'
 - en: Compile the model, using the AdamW optimizer (similar to Adam but with weight
     decay, which helps stabilize the training process) and mean absolute error loss
     function.
   id: totrans-205
   prefs: []
   type: TYPE_NORMAL
+  zh: 编译模型，使用AdamW优化器（类似于Adam，但带有权重衰减，有助于稳定训练过程）和平均绝对误差损失函数。
 - en: '[![3](Images/3.png)](#co_diffusion_models_CO8-3)'
   id: totrans-206
   prefs: []
   type: TYPE_NORMAL
+  zh: '[![3](Images/3.png)](#co_diffusion_models_CO8-3)'
 - en: Calculate the normalization statistics using the training set.
   id: totrans-207
   prefs: []