Function Differences with torch.nn.functional.binary_cross_entropy

torch.nn.functional.binary_cross_entropy

torch.nn.functional.binary_cross_entropy(
    input,
    target,
    weight=None,
    size_average=None,
    reduce=None,
    reduction='mean'
) -> Tensor

For more information, see torch.nn.functional.binary_cross_entropy.

mindspore.ops.binary_cross_entropy

mindspore.ops.binary_cross_entropy(
    logits,
    labels,
    weight=None,
    reduction='mean'
) -> Tensor

For more information, see mindspore.ops.binary_cross_entropy.

Differences

PyTorch: Compute the binary cross-entropy loss value between the target and predicted values.

MindSpore: MindSpore API basically implements the same function as PyTorch.

Categories	Subcategories	PyTorch	MindSpore	Differences
Parameters	Parameter 1	input	logits	Same function, different parameter names
	Parameter 2	target	labels	Same function, different parameter names
	Parameter 3	weight	weight	Same function
	Parameter 4	size_average	-	PyTorch deprecated parameters, functionally replaced by the reduction parameter
	Parameter 5	reduce	-	PyTorch deprecated parameters, functionally replaced by the reduction parameter
	Parameter 6	reduction	reduction	Same function

Code Example 1

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

# PyTorch
import torch
import torch.nn.functional as F
from torch import tensor

logits = tensor([0.1, 0.2, 0.3], requires_grad=True)
labels = tensor([1., 1., 1.])
loss = F.binary_cross_entropy(logits, labels)
print(loss.detach().numpy())
# 1.7053319

# MindSpore
import mindspore
import numpy as np
from mindspore import Tensor
from mindspore import ops

logits = Tensor(np.array([0.1, 0.2, 0.3]), mindspore.float32)
labels = Tensor(np.array([1., 1., 1.]), mindspore.float32)
loss = ops.binary_cross_entropy(logits, labels)
print(loss)
# 1.7053319