比较与torch.nn.MaxPool1d的功能差异

查看源文件

torch.nn.MaxPool1d

torch.nn.MaxPool1d(
    kernel_size,
    stride=None,
    padding=0,
    dilation=1,
    return_indices=False,
    ceil_mode=False
)

更多内容详见torch.nn.MaxPool1d

mindspore.nn.MaxPool1d

class mindspore.nn.MaxPool1d(
    kernel_size=1,
    stride=1,
    pad_mode="valid"
)

更多内容详见mindspore.nn.MaxPool1d

使用方式

PyTorch:可以通过padding参数调整输出的shape。若输入的shape为 \( (N, C, L_{in}) \),则输出的shape为 \( (N, C, L_{out}) \),其中

\[ L_{out} = \left\lfloor \frac{L_{in} + 2 \times \text{padding} - \text{dilation} \times (\text{kernel_size} - 1) - 1}{\text{stride}} + 1\right\rfloor \]

MindSpore:没有padding参数,仅通过pad_mode参数控制pad模式。若输入的shape为 \( (N, C, L_{in}) \),则输出的shape为 \( (N, C, L_{out}) \),其中

  1. pad_mode为”valid”:

    \[ L_{out} = \left\lfloor \frac{L_{in} - (\text{kernel_size} - 1)}{\text{stride}}\right\rfloor \]
  2. pad_mode为”same”:

    \[ L_{out} = \left\lfloor \frac{L_{in}}{\text{stride}}\right\rfloor \]

代码示例

import mindspore
from mindspore import Tensor
import mindspore.nn as nn
import torch
import numpy as np

# In MindSpore, pad_mode="valid"
pool = nn.MaxPool1d(kernel_size=3, stride=2, pad_mode="valid")
input_x = Tensor(np.random.randn(20, 16, 50).astype(np.float32))
output = pool(input_x)
print(output.shape)
# Out:
# (20, 16, 24)

# In MindSpore, pad_mode="same"
pool = nn.MaxPool1d(kernel_size=3, stride=2, pad_mode="same")
input_x = Tensor(np.random.randn(20, 16, 50).astype(np.float32))
output = pool(input_x)
print(output.shape)
# Out:
# (20, 16, 25)


# In torch, padding=1
m = torch.nn.MaxPool1d(3, stride=2, padding=1)
input_x = torch.randn(20, 16, 50)
output = m(input_x)
print(output.shape)
# Out:
# torch.Size([20, 16, 25])