[torch 参数更多 ] torch.nn.functional.triplet_margin_loss¶
torch.nn.functional.triplet_margin_loss¶
torch.nn.functional.triplet_margin_loss(anchor,
positive,
negative,
margin=1.0,
p=2,
eps=1e-06,
swap=False,
size_average=None,
reduce=None,
reduction='mean')
paddle.nn.functional.triplet_margin_loss¶
paddle.nn.functional.triplet_margin_loss(input,
positive,
negative,
margin: float = 1.0,
p: float = 2.0,
epsilon: float = 1e-6,
swap: bool = False,
reduction: str = 'mean',
name: str = None)
其中 Pytorch 相⽐ Paddle ⽀持更多其他参数,具体如下:
参数映射¶
| PyTorch | PaddlePaddle | 备注 | | ————- | ———— | —————————————————— | | anchor | input | 输入 Tensor,仅参数名不一致。 | | positive | positive | 输入正样本 | | negative | negative | 输入负样本 | | margin | margin | 手动指定间距 | | p | p | 指定范数 | | eps | epsilon | 防止除数为零的常数 | | swap | swap | 是否进行交换 | | size_average | - | 已弃用 | | reduce | - | 已弃用 | | reduction | reduction | 表示应用于输出结果的规约方式,可选值有:’none’, ‘mean’, ‘sum’ |