比较与torch.nn.ConvTranspose3d的差异
torch.nn.ConvTranspose3d
class torch.nn.ConvTranspose3d(
in_channels,
out_channels,
kernel_size,
stride=1,
padding=0,
output_padding=0,
groups=1,
bias=True,
dilation=1,
padding_mode='zeros'
)(input) -> Tensor
更多内容详见torch.nn.ConvTranspose3d。
mindspore.nn.Conv3dTranspose
class mindspore.nn.Conv3dTranspose(
in_channels,
out_channels,
kernel_size,
stride=1,
pad_mode='same',
padding=0,
dilation=1,
group=1,
output_padding=0,
has_bias=False,
weight_init=None,
bias_init=None,
data_format='NCDHW'
)(x) -> Tensor
更多内容详见mindspore.nn.Conv3dTranspose。
差异对比
PyTorch:计算三维转置卷积,可以视为Conv3d对输入求梯度,也称为反卷积(实际不是真正的反卷积)。输入的shape通常是\((N,C_{in},D_{in},H_{in},W_{in})\),其中\(N\)是batch size,\(C\)是空间维度,\(D_{in},H_{in},W_{in}\)分别为特征层的深度,高度和宽度。输出的shape为\((N,C_{out},D_{out},H_{out},W_{out})\),计算公式如下: \(D_{out}=(D_{in}−1)×stride[0]−2×padding[0]+dilation[0]×(kernel\underline{ }size[0]−1)+output\underline{ }padding[0]+1\) \(H_{out}=(H_{in}−1)×stride[1]−2×padding[1]+dilation[1]×(kernel\underline{ }size[1]−1)+output\underline{ }padding[1]+1\) \(W_{out}=(W_{in}−1)×stride[2]−2×padding[2]+dilation[2]×(kernel\underline{ }size[2]−1)+output\underline{ }padding[2]+1\)
MindSpore:MindSpore此API实现功能与PyTorch基本一致,新增了填充模式参数"pad_mode",当"pad_mode" = "pad"时与PyTorch默认方式相同,利用weight_init 和bias_init 参数可以配置初始化方式。
权重初始化差异
mindspore.nn.Conv3dTranspose的 weight_init
是 None
时,权重使用HeUniform初始化。此时和PyTorch权重初始化方式一致。
mindspore.nn.Conv3dTranspose的 bias_init
是 None
时,偏差使用Uniform初始化。此时和PyTorch偏差初始化方式一致。
分类 |
子类 |
PyTorch |
MindSpore |
差异 |
---|---|---|---|---|
参数 |
参数1 |
in_channels |
in_channels |
- |
参数2 |
out_channels |
out_channels |
- |
|
参数3 |
kernel_size |
kernel_size |
- |
|
参数4 |
stride |
stride |
- |
|
参数5 |
padding |
padding |
功能一致,PyTorch中只能在三个维度的两侧分别填充相同的值,可为长度为3的tuple。MindSpore中可以分别设置前部、尾部、顶部、底部、左边和右边的填充数量,可为长度为6的tuple |
|
参数6 |
output_padding |
output_padding |
- |
|
参数7 |
groups |
group |
功能一致,参数名不同 |
|
参数8 |
bias |
has_bias |
PyTorch默认为True,MindSpore默认为False |
|
参数9 |
dilation |
dilation |
- |
|
参数10 |
padding_mode |
- |
数值填充模式,只支持"zeros"即填充0。MindSpore无此参数,但默认填充0 |
|
参数11 |
- |
pad_mode |
指定填充模式。可选值为"same"、"valid"、"pad",在"same"和"valid"模式下,padding必须设置为0,默认为"same",PyTorch无此参数 |
|
参数12 |
- |
weight_init |
权重参数的初始化方法。可为Tensor,str,Initializer或numbers.Number。当使用str时,可选"TruncatedNormal","Normal","Uniform","HeUniform"和"XavierUniform"分布以及常量"One"和"Zero"分布的值。默认为None,PyTorch无此参数 |
|
参数13 |
- |
bias_init |
偏置参数的初始化方法。可选填参数与"weight_init"相同,默认为None,PyTorch无此参数 |
|
参数14 |
- |
data_format |
数据格式的可选值。目前仅支持"NCDHW",与PyTorch中默认顺序一致,PyTorch无此参数 |
|
输入 |
单输入 |
input |
x |
功能一致,参数名不同 |
代码示例1
两API都是实现三维转置卷积运算,使用时需先进行实例化。为使输出的宽度与输入整除stride后的值相同,PyTorch中设置output_padding = stride - 1,padding设置为(kernel_size - 1)/2。MindSpore则设置pad_mode = "same",同时padding = 0。
# PyTorch
import torch
from torch import tensor
import torch.nn as nn
import numpy as np
k = 5
s = 3
x_ = np.ones([1, 3, 4, 9, 16])
x = tensor(x_, dtype=torch.float32)
net = nn.ConvTranspose3d(3, 32, kernel_size=k, stride=s, padding=(k-1)//2, output_padding=s-1, bias=False)
net.weight.data = torch.ones(3, 32, k, k, k)
output = net(x).detach().numpy()
print(output.shape)
# (1, 32, 12, 27, 48)
# MindSpore
import mindspore as ms
import mindspore.nn as nn
import numpy as np
k = 5
s = 3
x_ = np.ones([1, 3, 4, 9, 16])
x = ms.Tensor(x_, ms.float32)
net = nn.Conv3dTranspose(3, 32, kernel_size=k, stride=s, weight_init='ones', pad_mode='same')
output = net(x)
print(output.shape)
# (1, 32, 12, 27, 48)
代码示例2
两API都是实现三维转置卷积运算,使用时需先进行实例化。若不在原有图像上做任何填充,在stride>1的情况下可能舍弃一部分数据,在PyTorch中将padding和output_padding设为0,MindSpore中设置pad_mode = "valid",同时padding = 0。
# PyTorch
import torch
from torch import tensor
import torch.nn as nn
import numpy as np
k = 5
s = 3
x_ = np.ones([1, 3, 4, 9, 16])
x = tensor(x_, dtype=torch.float32)
net = nn.ConvTranspose3d(3, 32, kernel_size=k, stride=s, bias=False)
net.weight.data = torch.ones(3, 32, k, k, k)
output = net(x).detach().numpy()
print(output.shape)
# (1, 32, 14, 29, 50)
# MindSpore
import mindspore as ms
import mindspore.nn as nn
import numpy as np
k = 5
s = 3
x_ = np.ones([1, 3, 4, 9, 16])
x = ms.Tensor(x_, ms.float32)
net = nn.Conv3dTranspose(3, 32, kernel_size=k, stride=s, weight_init='ones', pad_mode='valid')
output = net(x)
print(output.shape)
# (1, 32, 14, 29, 50)