update

1_5_CNN_Layers/4_3. Classify FashionMNIST, solution 2.ipynb

@@ -165,14 +165,14 @@
     "x = self.pool(F.relu(self.conv1(x)))\n",
     "```\n",
     "\n",
-    "Best practice is to place any layers whose weights will change during the training process in `__init__` and refer to them in the `forward` function; any layers or functions that always behave in the same way, such as a pre-defined activation function, should appear *only* in the `forward` function.\n",
+    "You must place any layers with trainable weights, such as convolutional layers, in the `__init__` function and refer to them in the `forward` function; any layers or functions that always behave in the same way, such as a pre-defined activation function, may appear *only* in the `forward` function. In practice, you'll often see conv/pool layers defined in `__init__` and activations defined in `forward`.\n",
     "\n",
     "#### Convolutional layer\n",
     "The first convolution layer has been defined for you, it takes in a 1 channel (grayscale) image and outputs 10 feature maps as output, after convolving the image with 3x3 filters.\n",
     "\n",
     "#### Flattening\n",
     "\n",
-    "Recall that to move from the output of a convoluttional/pooling layer to a linear layer, you must first flatten your extracted features into a vector. If you've used the deep learning library, Keras, you may have seen this done by `Flatten()`, and in PyTorch you can flatten an input `x` with `x = x.view(x.size(0), -1)`.\n",
+    "Recall that to move from the output of a convolutional/pooling layer to a linear layer, you must first flatten your extracted features into a vector. If you've used the deep learning library, Keras, you may have seen this done by `Flatten()`, and in PyTorch you can flatten an input `x` with `x = x.view(x.size(0), -1)`.\n",
     "\n",
     "### TODO: Define the rest of the layers\n",
     "\n",