Function Differences with torch.nn.AvgPool1d

View Source On Gitee

torch.nn.AvgPool1d

torch.nn.AvgPool1d(kernel_size, stride=None, padding=0, ceil_mode=False, count_include_pad=True)(input) -> Tensor

For more information, see torch.nn.AvgPool1d.

mindspore.nn.AvgPool1d

mindspore.nn.AvgPool1d(kernel_size=1, stride=1, pad_mode="valid", padding=0, ceil_mode=False, count_include_pad=True)(x) -> Tensor

For more information, see mindspore.nn.AvgPool1d.

Differences

PyTorch: Perform averaging pooling operations on a one-dimensional plane on the input multidimensional data.

MindSpore: This API implementation function of MindSpore is compatible with TensorFlow and PyTorch, When pad_mode is “valid” or “same”, the function is consistent with TensorFlow, and when pad_mode is “pad”, the function is consistent with PyTorch, MindSpore additionally supports 2D input, which is consistent with PyTorch 1.12.

Categories

Subcategories

PyTorch

MindSpore

Differences

Parameters

Parameter 1

kernel_size

kernel_size

Consistent function, no default values for PyTorch

Parameter 2

stride

stride

The functions are the same, but the default values of parameters are different

Parameter 3

padding

padding

Consistent

Parameter 4

ceil_mode

ceil_mode

Consistent

Parameter 5

count_include_pad

count_include_pad

Consistent

Parameter 6

-

pad_mode

MindSpore specifies how the pooling will be filled, with optional values of “same”, “valid” or “pad”. PyTorch does not have this parameter

Input

Single input

input

x

Interface input, same function, different parameter names

Code Example 1

The two APIs achieve the same function and have the same usage.

# PyTorch
import torch
import torch.nn as nn

m = nn.AvgPool1d(kernel_size=6, stride=1)
input_x = torch.tensor([[[1,2,3,4,5,6,7]]], dtype=torch.float32)
print(m(input_x).numpy())
# [[[3.5 4.5]]]

# MindSpore
import mindspore
import mindspore.nn as nn
from mindspore import Tensor

pool = nn.AvgPool1d(kernel_size=6, stride=1)
x = Tensor([[[1,2,3,4,5,6,7]]], dtype=mindspore.float32)
output = pool(x)
print(output)
# [[[3.5 4.5]]]

Code Example 2

Use pad mode to ensure functional consistency.

import torch
import mindspore.nn as nn
import mindspore.ops as ops

pool = nn.AvgPool1d(4, stride=1, ceil_mode=True, pad_mode='pad', padding=2)
x1 = ops.randn(6, 6, 8)
output = pool(x1)
print(output.shape)
# (6, 6, 9)

pool = torch.nn.AvgPool1d(4, stride=1, ceil_mode=True, padding=2)
x1 = torch.randn(6, 6, 8)
output = pool(x1)
print(output.shape)
# torch.Size([6, 6, 9])