\u200E

[ 仅 API 调用方式不一致 ]torch.nn.HuberLoss

torch.nn.HuberLoss 

          torch.nn.HuberLoss(reduction='mean', delta=1.0)

         

paddle.nn.SmoothL1Loss 

          paddle.nn.SmoothL1Loss(reduction='mean', delta=1.0, name=None)

         

两者功能一致，但调用方式不一致，具体如下：

转写示例

          # PyTorch 写法
loss = torch.nn.HuberLoss()

# Paddle 写法
loss = paddle.nn.SmoothL1Loss()