mindspore.ops.SmoothL1Loss

class mindspore.ops.SmoothL1Loss(*args, **kwargs)

Computes smooth L1 loss, a robust L1 loss.

SmoothL1Loss is a Loss similar to MSELoss but less sensitive to outliers as described in the Fast R-CNN by Ross Girshick.

The updating formulas of SmoothL1Loss algorithm are as follows,

$$\text{SmoothL1Loss}=\{\begin{array}{ll}\frac{0.5x^2}{\text{beta}},& if\left|x\right|<\text{beta}\\ \left|x\right|-0.5\text{beta},& \text{otherwise}\end{array}$$

where $X$ represents prediction. $Y$ represents target. $loss$ represents output.

Parameters

beta (float) – A parameter used to control the point where the function will change from quadratic to linear. Default: 1.0.

Inputs:

• prediction (Tensor) - Predict data. Data type must be float16 or float32.

• target (Tensor) - Ground truth data, with the same type and shape as prediction.

Outputs:

Tensor, with the same type and shape as prediction.

Raises

• TypeError – If beta is not a float.

• TypeError – If prediction or target is not a Tensor.

• TypeError – If dtype of prediction or target is neither float16 nor float32.

• ValueError – If beta is less than or equal to 0.

• ValueError – If shape of prediction is not the same as target.

Supported Platforms:

Ascend GPU CPU

Examples

>>> loss = ops.SmoothL1Loss()
>>> input_data = Tensor(np.array([1, 2, 3]), mindspore.float32)
>>> target_data = Tensor(np.array([1, 2, 2]), mindspore.float32)
>>> output = loss(input_data, target_data)
>>> print(output)
[0.  0.  0.5]