Function Differences with torch.nn.BatchNorm1d

torch.nn.BatchNorm1d

class torch.nn.BatchNorm1d(
    num_features,
    eps=1e-05,
    momentum=0.1,
    affine=True,
    track_running_stats=True
)(input) -> Tensor

For more information, see torch.nn.BatchNorm1d.

mindspore.nn.BatchNorm1d

class mindspore.nn.BatchNorm1d(
    num_features,
    eps=1e-5,
    momentum=0.9,
    affine=True,
    gamma_init='ones',
    beta_init='zeros',
    moving_mean_init='zeros',
    moving_var_init='ones',
    use_batch_statistics=None,
    data_format='NCHW'
)(x) -> Tensor

For more information, see mindspore.nn.BatchNorm1d.

Differences

PyTorch：Batch normalization of the input 2D or 3D data.

MindSpore：The implementation function of the API in MindSpore is basically the same as that of PyTorch. The default value of the momentum parameter in MindSpore is 0.9, and the momentum conversion relationship with PyTorch is 1-momentum. The behavior of the default value is the same as that of PyTorch. The parameter update strategy during training and inference is different from that of PyTorch. For details, please refer to Differences Between MindSpore and PyTorch - nn.BatchNorm2d.

Categories	Subcategories	PyTorch	MindSpore	Differences
Parameters	Parameter 1	num_features	num_features	-
	Parameter 2	eps	eps	-
	Parameter 3	momentum	momentum	The function is the same, but the default value in PyTorch is 0.1, and in MindSpore is 0.9. The conversion relationship with PyTorch’s momentum is 1-momentum, and the default value behavior is the same as PyTorch
	Parameter 4	affine	affine	-
	Parameter 5	track_running_stats	use_batch_statistics	The function is the same, and different values correspond to different default methods. For details, please refer to Typical differences with PyTorch - BatchNorm
	Parameter 6	-	gamma_init	PyTorch does not have this parameter, while MindSpore can initialize the value of the parameter gamma
	Parameter 7	-	beta_init	PyTorch does not have this parameter, while MindSpore can initialize the value of the parameter beta
	Parameter 8	-	moving_mean_init	PyTorch does not have this parameter, while MindSpore can initialize the value of the parameter moving_mean
	Parameter 9	-	moving_var_init	PyTorch does not have this parameter, while MindSpore can initialize the value of the parameter moving_var
	Parameter 10	-	data_format	PyTorch does not have this parameter
Input	Single input	input	x	Same function, different parameter names

Code Example

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

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

net = nn.BatchNorm1d(4, affine=False, momentum=0.1)
x = tensor(np.array([[0.7, 0.5, 0.5, 0.6], [0.5, 0.4, 0.6, 0.9]]).astype(np.float32))
output = net(x)
print(output.detach().numpy())
# [[ 0.9995001   0.9980063  -0.998006   -0.99977785]
#  [-0.9995007  -0.9980057   0.998006    0.99977785]]

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

net = nn.BatchNorm1d(num_features=4, affine=False, momentum=0.9)
net.set_train()
# BatchNorm1d<num_features=4, eps=1e-05, momentum=0.9, gamma=Parameter (name=gamma, shape=(4,), dtype=Float32, requires_grad=False), beta=Parameter (name=beta, shape=(4,), dtype=Float32, requires_grad=False), moving_mean=Parameter (name=mean, shape=(4,), dtype=Float32, requires_grad=False), moving_variance=Parameter (name=variance, shape=(4,), dtype=Float32, requires_grad=False)>

x = Tensor(np.array([[0.7, 0.5, 0.5, 0.6], [0.5, 0.4, 0.6, 0.9]]).astype(np.float32))
output = net(x)
print(output.asnumpy())
# [[ 0.9995001  0.9980063 -0.998006  -0.9997778]
#  [-0.9995007 -0.9980057  0.998006   0.9997778]]