Differences between torch.nn.AdaptiveAvgPool1d

The following mapping relationships can be found in this file.

PyTorch APIs	MindSpore APIs
torch.nn.AdaptiveAvgPool1d	mindspore.nn.AdaptiveAvgPool1d
torch.nn.functional.adaptive_avg_pool1d	mindspore.ops.adaptive_avg_pool1d

torch.nn.AdaptiveAvgPool1d

torch.nn.AdaptiveAvgPool1d(output_size)(input) -> Tensor

For more information, see torch.nn.AdaptiveAvgPool1d.

mindspore.nn.AdaptiveAvgPool1d

mindspore.nn.AdaptiveAvgPool1d(output_size)(input) -> Tensor

For more information, see mindspore.nn.AdaptiveAvgPool1d.

Differences

PyTorch: Adaptive average pooling operation for temporal data, supporting 2D and 3D data.

MindSpore: This API in MindSpore currently only supports 3D data and requires the last dimension of the input data to be larger than that of the output data, and must divide the output_size.

Categories	Subcategories	PyTorch	MindSpore	Differences
Parameters	Parameter 1	output_size	output_size	MindSpore requires the last dimension of the input data to be larger than that of the output data, and must divide the output_size.
Input	Single input	input	input	MindSpore currently only supports 3D data

Code Example 1

For 3D data, perform adaptive average pooling operation on the data when the output length can be divisiable by the input length.

# PyTorch
import torch
from torch import tensor
import numpy as np

avg_pool = torch.nn.AdaptiveAvgPool1d(output_size=4)
x = tensor(np.arange(16).reshape(1, 2, 8), dtype=torch.float32)
output = avg_pool(x)
print(output)
# tensor([[[ 0.5000,  2.5000,  4.5000,  6.5000],
#          [ 8.5000, 10.5000, 12.5000, 14.5000]]])

# MindSpore
import mindspore
from mindspore import Tensor, nn
import numpy as np
pool = nn.AdaptiveAvgPool1d(output_size=4)
x = Tensor(np.arange(16).reshape(1, 2, 8), mindspore.float32)
output = pool(x)
print(output)
# [[[ 0.5  2.5  4.5  6.5]
#   [ 8.5 10.5 12.5 14.5]]]