DEV Community

Super Kai (Kazuya Ito)
Super Kai (Kazuya Ito)

Posted on • Edited on

L1Loss and MSELoss in PyTorch

Buy Me a Coffee

*Memos:

L1Loss() can get the 0D or more D tensor of the zero or more values(float) computed by L1 Loss(MAE) from the 0D or more D tensor of zero or more elements as shown below:

*Memos:

  • There is reduction argument for initialization(Optional-Default:'mean'-Type:str). *'none', 'mean' or 'sum' can be selected.
  • There are size_average and reduce argument for initialization but they are deprecated.
  • The 1st argument is input(Required-Type:tensor of float or complex).
  • The 2nd argument is target(Required-Type:tensor of float or complex).
  • input and target should be the same size otherwise there is a warning.
  • Even complex type of input and target tensors return a float tensor.
  • The empty 1D or more D input and target tensor with reduction='mean' return nan.
  • The empty 1D or more D input and target tensor with reduction='sum' return 0.. Image description
import torch
from torch import nn

tensor1 = torch.tensor([ 8., -3., 0.,  1.,  5., -2., -1., 4.])
tensor2 = torch.tensor([-3.,  7., 4., -2., -9.,  6., -8., 5.])
                      # |x-y|
                      # |8.-(-3.)| = 11.
                      # ↓↓
                      # 11.+ 10.+ 4. + 3.+ 14. + 8. + 7.+ 1. = 58.
                      # 58. / 8 = 7.25
l1loss = nn.L1Loss()
l1loss(input=tensor1, target=tensor2)
# tensor(7.2500)

l1loss
# L1Loss()

l1loss.reduction
# 'mean'

l1loss = nn.L1Loss(reduction='mean')
l1loss(input=tensor1, target=tensor2)
# tensor(7.2500)

l1loss = nn.L1Loss(reduction='sum')
l1loss(input=tensor1, target=tensor2)
# tensor(58.)

l1loss = nn.L1Loss(reduction='none')
l1loss(input=tensor1, target=tensor2)
# tensor([11., 10., 4., 3., 14., 8., 7., 1.])

tensor1 = torch.tensor([[8., -3., 0., 1.], [5., -2., -1., 4.]])
tensor2 = torch.tensor([[-3., 7., 4., -2.], [-9., 6., -8., 5.]])

l1loss = nn.L1Loss()
l1loss(input=tensor1, target=tensor2)
# tensor(7.2500)

tensor1 = torch.tensor([[[8., -3.], [0., 1.]], [[5., -2.], [-1., 4.]]])
tensor2 = torch.tensor([[[-3., 7.], [4., -2.]], [[-9., 6.], [-8., 5.]]])

l1loss = nn.L1Loss()
l1loss(input=tensor1, target=tensor2)
# tensor(7.2500)

tensor1 = torch.tensor([[[8.+0.j, -3.+0.j], [0.+0.j, 1.+0.j]],
                        [[5.+0.j, -2.+0.j], [-1.+0.j, 4.+0.j]]])
tensor2 = torch.tensor([[[-3.+0.j, 7.+0.j], [4.+0.j, -2.+0.j]],
                        [[-9.+0.j, 6.+0.j], [-8.+0.j, 5.+0.j]]])
l1loss = nn.L1Loss()
l1loss(input=tensor1, target=tensor2)
# tensor(7.2500)

tensor1 = torch.tensor([])
tensor2 = torch.tensor([])

l1loss = nn.L1Loss(reduction='mean')
l1loss(input=tensor1, target=tensor2)
# tensor(nan)

l1loss = nn.L1Loss(reduction='sum')
l1loss(input=tensor1, target=tensor2)
# tensor(0.)
Enter fullscreen mode Exit fullscreen mode

MSELoss() can get the 0D or more D tensor of the zero or more values(float) computed by L2 Loss(MSE) from the 0D or more D tensor of zero or more elements as shown below:

*Memos:

  • There is reduction argument for initialization(Optional-Default:'mean'-Type:str). *'none', 'mean' or 'sum' can be selected.
  • There are size_average and reduce argument for initialization but they are deprecated.
  • The 1st argument is input(Required-Type:tensor of float).
  • The 2nd argument is target(Required-Type:tensor of float).
  • input and target should be the same size otherwise there is a warning.
  • The empty 1D or more D input and target tensor with reduction='mean' return nan.
  • The empty 1D or more D input and target tensor with reduction='sum' return 0.. Image description
import torch
from torch import nn

tensor1 = torch.tensor([ 8., -3., 0.,  1.,  5., -2., -1., 4.])
tensor2 = torch.tensor([-3.,  7., 4., -2., -9.,  6., -8., 5.])
                     # (x-y)^2
                     # (8.-(-3.))^2 = 121.
                     # ↓↓↓
                     # 121. 100. 16.   9. 196.  64.  49.  1. = 556.
                     # 556. / 8 = 69.5
mseloss = nn.MSELoss()
mseloss(input=tensor1, target=tensor2)
# tensor(69.5000)

mseloss
# MSELoss()

mseloss.reduction
# 'mean'

mseloss = nn.MSELoss(reduction='mean')
mseloss(input=tensor1, target=tensor2)
# tensor(69.5000)

mseloss = nn.MSELoss(reduction='sum')
mseloss(input=tensor1, target=tensor2)
# tensor(556.)

mseloss = nn.MSELoss(reduction='none')
mseloss(input=tensor1, target=tensor2)
# tensor([121., 100., 16., 9., 196., 64., 49., 1.])

tensor1 = torch.tensor([[8., -3., 0., 1.], [5., -2., -1., 4.]])
tensor2 = torch.tensor([[-3., 7., 4., -2.], [-9., 6., -8., 5.]])

mseloss = nn.MSELoss()
mseloss(input=tensor1, target=tensor2)
# tensor(69.5000)

tensor1 = torch.tensor([[[8., -3.], [0., 1.]], [[5., -2.], [-1., 4.]]])
tensor2 = torch.tensor([[[-3., 7.], [4., -2.]], [[-9., 6.], [-8., 5.]]])

mseloss = nn.MSELoss()
mseloss(input=tensor1, target=tensor2)
# tensor(69.5000)

tensor1 = torch.tensor([])
tensor2 = torch.tensor([])

mseloss = nn.MSELoss(reduction='mean')
mseloss(input=tensor1, target=tensor2)
# tensor(nan)

mseloss = nn.MSELoss(reduction='sum')
mseloss(input=tensor1, target=tensor2)
# tensor(0.)
Enter fullscreen mode Exit fullscreen mode

Heroku

Build apps, not infrastructure.

Dealing with servers, hardware, and infrastructure can take up your valuable time. Discover the benefits of Heroku, the PaaS of choice for developers since 2007.

Visit Site

Top comments (0)

A Workflow Copilot. Tailored to You.

Pieces.app image

Our desktop app, with its intelligent copilot, streamlines coding by generating snippets, extracting code from screenshots, and accelerating problem-solving.

Read the docs

👋 Kindness is contagious

Discover a treasure trove of wisdom within this insightful piece, highly respected in the nurturing DEV Community enviroment. Developers, whether novice or expert, are encouraged to participate and add to our shared knowledge basin.

A simple "thank you" can illuminate someone's day. Express your appreciation in the comments section!

On DEV, sharing ideas smoothens our journey and strengthens our community ties. Learn something useful? Offering a quick thanks to the author is deeply appreciated.

Okay