Created by: still-wait

PR types

New features

PR changes

APIs

Describe

add paddle.nn.SmoothL1Loss. This operator is calculate smooth_l1_loss. Creates a criterion that uses a squared term if the absolute element-wise error falls below 1 and an L1 term otherwise. In some cases it can prevent exploding gradients. Also known as the Huber loss.

class:

class paddle.nn.SmoothL1Loss(reduction='mean')

functioanl:

def paddle.nn.functioanl.smooth_l1_loss(input, label)

docs

Examples:

           # declarative mode
            import paddle
            import paddle.fluid as fluid
            import numpy as np
            input = fluid.layers.data(name="input", shape=[-1, 3], dtype="float32")
            label = fluid.layers.data(name="label", shape=[-1, 3], dtype="float32")
            loss = paddle.nn.SmoothL1Loss()
            result = loss(x,label)
            place = fluid.CPUPlace()
            exe = fluid.Executor(place)
            exe.run(fluid.default_startup_program())
            input_data = np.random.rand(3,3).astype("float32")
            label_data = np.random.rand(3,3).astype("float32")
            output= exe.run(feed={"input": input_data, "label": label_data},
                            fetch_list=[result])
            print(output)

            # imperative mode
            import paddle.fluid.dygraph as dg
            with dg.guard(place) as g:
                input = dg.to_variable(input_data)
                label = dg.to_variable(label_data)
                weight = dg.to_variable(weight_data)
                loss = paddle.nn.SmoothL1Loss()
                output = loss(input, label)
                print(output.numpy())