smooth_l1_loss¶
dragon.losses.
smooth_l1_loss
(
inputs,
beta=1.0,
reduction='mean',
**kwargs
)[source]¶Compute the loss of element-wise error transited from L1 and L2. [Girshick, 2015].
The SmoothL1Loss function is defined as:
\[\text{SmoothL1Loss}(x, y) = \begin{cases} 0.5 * (x - y)^{2} / beta, & \text{ if } |x - y| < beta \\ |x - y| - 0.5 * beta, & \text{ otherwise } \end{cases} \]Examples:
x = dragon.constant([1., 2., 3.]) y = dragon.constant([0., 0., 0.]) print(dragon.losses.smooth_l1_loss([x, y])) # 1.5
- Parameters:
- inputs (Sequence[dragon.Tensor]) – The tensor
input
andtarget
. - beta (float, optional, default=1.0) – The transition point from L1 to L2 loss
- reduction ({'none', 'sum', 'mean'}, optional) – The reduction method.
- inputs (Sequence[dragon.Tensor]) – The tensor
- Returns:
dragon.Tensor – The output tensor.