[ torch 参数更多 ]torch.nn.MultiLabelSoftMarginLoss

torch.nn.MultiLabelSoftMarginLoss

torch.nn.MultiLabelSoftMarginLoss(weight=None, size_average=None, reduce=None, reduction='mean')

paddle.nn.MultiLabelSoftMarginLoss

paddle.nn.MultiLabelSoftMarginLoss(weight=None, reduction='mean', name=None)

PyTorch 相比 Paddle 支持更多其他参数，具体如下：

参数映射

PyTorch PaddlePaddle 备注
weight weight 手动设定权重。
size_average - 已废弃，和 reduce 组合决定损失计算方式。需要转写。
reduce - 已废弃，和 size_average 组合决定损失计算方式。需要转写。
reduction reduction 指定应用于输出结果的计算方式。

转写示例

size_average/reduce：对应到 reduction 为 sum

# PyTorch 写法
torch.nn.MultiLabelSoftMarginLoss(weight=w, size_average=False, reduce=True)
torch.nn.MultiLabelSoftMarginLoss(weight=w, size_average=False)

# Paddle 写法
paddle.nn.MultiLabelSoftMarginLoss(weight=w, reduction='sum')

size_average/reduce：对应到 reduction 为 mean

# PyTorch 写法
torch.nn.MultiLabelSoftMarginLoss(weight=w, size_average=True, reduce=True)
torch.nn.MultiLabelSoftMarginLoss(weight=w, reduce=True)
torch.nn.MultiLabelSoftMarginLoss(weight=w, size_average=True)
torch.nn.MultiLabelSoftMarginLoss(weight=w)

# Paddle 写法
paddle.nn.MultiLabelSoftMarginLoss(weight=w, reduction='mean')

size_average/reduce：对应到 reduction 为 none

# PyTorch 写法
torch.nn.MultiLabelSoftMarginLoss(weight=w, size_average=True, reduce=False)
torch.nn.MultiLabelSoftMarginLoss(weight=w, size_average=False, reduce=False)
torch.nn.MultiLabelSoftMarginLoss(weight=w, reduce=False)

# Paddle 写法
paddle.nn.MultiLabelSoftMarginLoss(weight=w, reduction='none')