Function Differences with torch.nn.CosineEmbeddingLoss

torch.nn.CosineEmbeddingLoss

class torch.nn.CosineEmbeddingLoss(
    margin=0.0,
    size_average=None,
    reduce=None,
    reduction='mean'
)(x1, x2, target) -> Tensor

For more information, see torch.nn.CosineEmbeddingLoss.

mindspore.nn.CosineEmbeddingLoss

class mindspore.nn.CosineEmbeddingLoss(
    margin=0.0,
    reduction='mean'
)(logits_x1, logits_x2, labels) -> Tensor/Scalar

For more information, see mindspore.nn.CosineEmbeddingLoss.

Differences

PyTorch: Cosine similarity loss function for measuring the similarity between two Tensors.

MindSpore: Implement the same function as PyTorch.

Categories	Subcategories	PyTorch	MindSpore	Difference
Parameters	Parameter 1	margin	margin	-
	Parameter 2	size_average	-	Deprecated, function taken over by reduction
	Parameter 3	reduce	-	Deprecated, function taken over by reduction
	Parameter 4	reduction	reduction	-
Input	Input 1	x1	logits_x1	Same function, different parameter names
	Input 2	x2	logits_x2	Same function, different parameter names
	Input 3	target	labels	Same function, different parameter names

Code Example

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

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

input1 = tensor(np.array([[0.3, 0.8], [0.4, 0.3]]), dtype=torch.float32)
input2 = tensor(np.array([[0.4, 1.2], [-0.4, -0.9]]), dtype=torch.float32)
target = tensor(np.array([1, -1]), dtype=torch.int32)
cosine_embedding_loss = nn.CosineEmbeddingLoss()
output = cosine_embedding_loss(input1, input2, target)
print(output.detach().numpy())
# 0.0003426075

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

logits_x1 = Tensor(np.array([[0.3, 0.8], [0.4, 0.3]]), mindspore.float32)
logits_x2 = Tensor(np.array([[0.4, 1.2], [-0.4, -0.9]]), mindspore.float32)
labels = Tensor(np.array([1, -1]), mindspore.int32)
cosine_embedding_loss = nn.CosineEmbeddingLoss()
output = cosine_embedding_loss(logits_x1, logits_x2, labels)
print(output)
# 0.0003425479