Today, December 8th, 2018, PyTorch 1.0 stable has been released. It is a milestone and I’d like to keep notes on PyTorch as I learn and use PyTorch. The resource mainly comes from PyTorch official tutorial and Intro to Deep Learning with PyTorch on Udacity.
TENSORS are a generalization of vectors and matrices. In PyTorch, they are a multi-dimensional matrix containing elements of a single data type.
# From the slide PyTorch under the hood
Resizing the tensor
There are a few options to use:
w.reshape(a, b)will return a new tensor with the same data as
(a, b)sometimes, and sometimes a clone, as in it copies the data to another part of memory
w.resize_(a, b)returns the same tensor with a different shape. However, if the new shape results in fewer elements than the original tensor, some elements will be removed from the tensor (but not from memory). If the new shape results in more elements than the original tensor, new elements will be uninitialized in memory.
w.view(a, b)will return a new tensor with the same data as
The above three methods are introduced in Intro to Deep Learning with PyTorch. PyTorch official tutorial only introduces
w.view(), so generally I’d use
w.view() for resizing.
Convenience of -1
When resizing the tensor,
-1 is helpful to determine the only unknown size when we already know the other sizes. It can be inferred from other dimensions.
x = torch.randn(4, 4)
The size of y is
torch.Size([2, 8]), just as we want.
An in-place operation is an operation that changes directly the content of a given tensor without making a copy. In-place operations in PyTorch are always postfixed with a
.resize_() mentioned above is also an in-place operation.
NumPy to Torch and back
PyTorch has a great feature for converting between NumPy arrays and Torch tensors. To create a tensor from a NumPy array, use
torch.from_numpy(). To convert a tensor to a NumPy array, use the
import numpy as np
The memory is shared between the NumPy array and Torch tensor.
dim=0 takes the sum across the rows, i.e., compute the sum of the column vector. Similarly,
dim=1 takes the sum across the columns (compute the sum of the row vector).
The general process with PyTorch:
- Make a forward pass through the network
- Use the network output to calculate the loss
- Perform a backward pass through the network with
loss.backward()to calculate the gradients
- Take a step with the optimizer to update the weights
Constructing Neural Networks
import torch.nn as nn
It is mandatory to inherit from
nn.Module when creating a class for our network. The name of the class itself can be anything.
PyTorch networks created with
nn.Module must have a
forward method defined. It takes in a tensor
x and passes it through the operations you defined in the
__init__ method. And the
backward function (where gradients are computed) is automatically defined for you using
Another way is mentioned in the course:
nn.Sequential. (See Doc in detail)
model = nn.Sequential(nn.Linear(input_size, hidden_sizes),
PyTorch provides losses such as the cross-entropy loss (
nn.CrossEntropyLoss). You’ll usually see the loss assigned to
criterion. This criterion combines
nn.NLLLoss() in one single class.
The input is expected to contain scores for each class.
loss = nn.CrossEntropyLoss()
autograd package provides automatic differentiation for all operations on Tensors. If the attribute
torch.Tensor is set as
True, it starts to track all operations on it. When you finished your computation you can call
.backward() and have all the gradients computed automatically. The gradient for this tensor will be accumulated into
The most common method to reduce overfitting is dropout, where we randomly drop input units. Adding dropout in PyTorch is straightforward using the
self.dropout = nn.Dropout(p=0.2)
Turning off gradient descent when tesing or validating can help accelerate the process. Generally we can do the testing in the following mode:
During training we want to use dropout to prevent overfitting, but during inference we want to use the entire network. So, we need to turn off dropout during validation, testing, and whenever we’re using the network to make predictions. To do this, you use
model.eval(). This sets the model to evaluation mode where the dropout probability is 0. You can turn dropout back on by setting the model to train mode with
We typically define a convolutional layer in PyTorch using
nn.Conv2d(in_channels, out_channels, kernel_size, stride=1, padding=0)
in_channelsnumber of channels in the input image. For a grayscale image, this depth = 1; for a RGB image, this depth = 3.
out_channelsnumber of channels produced by the convolution
kernel_sizesize of the convolutional kernel (most commonly 3 for a 3x3 kernel)
stridestride of the convolution (default: 1, a single number or a tuple)
paddingzero-padding added to both sides of the input (default: 0, a number or a tuple)
kernel_sizethe size of the window to take a max over
stridethe stride of the window (default value is