ConvTranspose3d in PyTorch

Buy Me a Coffee☕

*Memos:

• My post explains Transposed Convolutional Layer.
• My post explains ConvTranspose1d().
• My post explains ConvTranspose2d().
• My post explains manual_seed().
• My post explains requires_grad.

ConvTranspose3d() can get the 4D or 5D tensor of the one or more elements computed by 3D transposed convolution from the 4D or 5D tensor of one or more elements as shown below:

*Memos:

• The 1st argument for initialization is in_channels(Required-Type:float). *It must be 1 <= x.
• The 2nd argument for initialization is out_channels(Required-Type:float). *It must be 1 <= x.
• The 3rd argument for initialization is kernel_size(Required-Type:int or tuple or list of int). *It must be 1 <= x.
• The 4th argument for initialization is stride(Optional-Default:1-Type:int or tuple or list of int). *It must be 1 <= x.
• The 5th argument for initialization is padding(Optional-Default:0-Type:int or tuple or list of int). *It must be 0 <= x.
• The 6th argument for initialization is output_padding(Optional-Default:0-Type:int, tuple or list of int). *It must be 0 <= x.
• The 7th argument for initialization is groups(Optional-Default:1-Type:int). *It must be 1 <= x.
• The 8th argument for initialization is bias(Optional-Default:True-Type:bool). *My post explains bias argument.
• The 9th argument for initialization is dilation(Optional-Default:1-Type:int or tuple or list of int). *It must be 1 <= x.
• The 10th argument for initialization is padding_mode(Optional-Default:'zeros'-Type:str). *Only 'zeros' can be selected.
• The 11th argument for initialization is device(Optional-Default:None-Type:str, int or device()): *Memos:
  • If it's None, get_default_device() is used. *My post explains get_default_device() and set_default_device().
  • device= can be omitted.
  • My post explains device argument.
• The 12th argument for initialization is dtype(Optional-Default:None-Type:dtype): *Memos:
  • If it's None, get_default_dtype() is used. *My post explains get_default_dtype() and set_default_dtype().
  • dtype= can be omitted.
  • My post explains dtype argument.
• The 1st argument is input(Required-Type:tensor of float or complex): *Memos:
  • It must be the 4D or 5D tensor of one or more elements.
  • The number of the elements of the 4th deepest dimension must be same as in_channels.
  • Its device and dtype must be same as ConvTranspose3d()'s.
  • complex must be set to dtype of ConvTranspose3d() to use a complex tensor.
  • The tensor's requires_grad which is False by default is set to True by ConvTranspose3d().
• convtran3d1.device and convtran3d1.dtype don't work.

import torch
from torch import nn

tensor1 = torch.tensor([[[[8., -3., 0., 1., 5., -2.]]]])

tensor1.requires_grad
# False

torch.manual_seed(42)

convtran3d1 = nn.ConvTranspose3d(in_channels=1, out_channels=3, kernel_size=1)
tensor2 = convtran3d1(input=tensor1)
tensor2
# tensor([[[[-2.6722, 0.4197, -0.4236, -0.7046, -1.8290, 0.1386]]],
#         [[[3.2144, -0.5154, 0.5018, 0.8409, 2.1972, -0.1763]]],
#         [[[4.1797, -1.4188, 0.1081, 0.6170, 2.6529, -0.9099]]]],
#        grad_fn=<SqueezeBackward1>)

tensor2.requires_grad
# True

convtran3d1
# ConvTranspose3d(1, 3, kernel_size=(1, 1, 1), stride=(1, 1, 1))

convtran3d1.in_channels
# 1

convtran3d1.out_channels
# 3

convtran3d1.kernel_size
# (1, 1, 1)

convtran3d1.stride
# (1, 1, 1)

convtran3d1.padding
# (0, 0, 0)

convtran3d1.output_padding
# (0, 0, 0)

convtran3d1.groups
# 1

convtran3d1.bias
# Parameter containing:
# tensor([0.5304, -0.1265, 0.1165], requires_grad=True)

convtran3d1.dilation
# (1, 1, 1)

convtran3d1.padding_mode
# 'zeros'

convtran3d1.weight
# Parameter containing:
# tensor([[[[[0.4414]]], [[[0.4792]]], [[[-0.1353]]]]], requires_grad=True)

torch.manual_seed(42)

convtran3d2 = nn.ConvTranspose3d(in_channels=3, out_channels=3, kernel_size=1)
convtran3d2(input=tensor2)
# tensor([[[[3.6068, -1.7503, -0.2893, 0.1977, 2.1458, -1.2633]]],
#         [[[1.6518, 0.4964, 0.8115, 0.9165, 1.3367, 0.6014]]],
#         [[[-0.5008, 0.2990, 0.0809, 0.0082, -0.2827, 0.2263]]]],
#        grad_fn=<SqueezeBackward1>)

torch.manual_seed(42)

convtran3d = nn.ConvTranspose3d(in_channels=1, out_channels=3, 
             kernel_size=1, stride=1, padding=0, output_padding=0, 
             groups=1, bias=True, dilation=1, padding_mode='zeros', 
             device=None, dtype=None)
convtran3d(input=tensor1)
# tensor([[[[4.0616, -0.7939, 0.5304, 0.9718, 2.7374, -0.3525]]],
#         [[[3.7071, -1.5641, -0.1265, 0.3527, 2.2695, -1.0849]]],
#         [[[-0.9656, 0.5223, 0.1165, -0.0188, -0.5598, 0.3870]]]],
#        grad_fn=<SqueezeBackward1>)

my_tensor = torch.tensor([[[[8., -3., 0.],
                            [1., 5., -2.]]]])
torch.manual_seed(42)

convtran3d = nn.ConvTranspose3d(in_channels=1, out_channels=3,
                                kernel_size=1)
convtran3d(input=my_tensor)
# tensor([[[[4.0616, -0.7939, 0.5304], [0.9718, 2.7374, -0.3525]]],
#         [[[3.7071, -1.5641, -0.1265], [0.3527, 2.2695, -1.0849]]],
#         [[[-0.9656, 0.5223, 0.1165], [-0.0188, -0.5598, 0.3870]]]],
#        grad_fn=<SqueezeBackward1>)

my_tensor = torch.tensor([[[[8.], [-3.], [0.],
                            [1.], [5.], [-2.]]]])
torch.manual_seed(42)

convtran3d = nn.ConvTranspose3d(in_channels=1, out_channels=3, kernel_size=1)
convtran3d(input=my_tensor)
# tensor([[[[4.0616], [-0.7939], [0.5304], [0.9718], [2.7374], [-0.3525]]],
#         [[[3.7071], [-1.5641], [-0.1265], [0.3527], [2.2695], [-1.0849]]],
#         [[[-0.9656], [0.5223], [0.1165], [-0.0188], [-0.5598], [0.3870]]]],
#        grad_fn=<SqueezeBackward1>)

my_tensor = torch.tensor([[[[[8.], [-3.], [0.]],
                            [[1.], [5.], [-2.]]]]])
torch.manual_seed(42)

convtran3d = nn.ConvTranspose3d(in_channels=1, out_channels=3, kernel_size=1)
convtran3d(input=my_tensor)
# tensor([[[[[4.0616], [-0.7939], [0.5304]],
#           [[0.9718], [2.7374], [-0.3525]]],
#          [[[3.7071], [-1.5641], [-0.1265]],
#           [[0.3527], [2.2695], [-1.0849]]],
#          [[[-0.9656], [0.5223], [0.1165]],
#           [[-0.0188], [-0.5598], [0.3870]]]]],
#        grad_fn=<ConvolutionBackward0>)

my_tensor = torch.tensor([[[[[8.+0.j], [-3.+0.j], [0.+0.j]],
                            [[1.+0.j], [5.+0.j], [-2.+0.j]]]]])
torch.manual_seed(42)

convtran3d = nn.ConvTranspose3d(in_channels=1, out_channels=3,
                                kernel_size=1, dtype=torch.complex64)
convtran3d(input=my_tensor)
# tensor([[[[[3.2502+4.1727j], [-1.6053-1.0985j], [-0.2811+0.3391j]],
#           [[0.1603+0.8183j], [1.9259+2.7351j], [-1.1639-0.6193j]]],
#          [[[-0.5731+3.8193j], [0.9147-2.0146j], [0.5090-0.4236j]],
#           [[0.3737+0.1068j], [-0.1673+2.2282j], [0.7795-1.4843j]]],
#          [[[-0.5102+1.0401j], [0.8813-0.2415j], [0.5018+0.1081j]],
#           [[0.3753+0.2246j], [-0.1307+0.6906j], [0.7548-0.1250j]]]]],
#        grad_fn=<AddBackward0>)