mindspore.ops.Dropout2D

class mindspore.ops.Dropout2D(keep_prob=0.5)[source]

During training, randomly zeroes some of the channels of the input tensor with probability 1-keep_prob from a Bernoulli distribution(For a 4-dimensional tensor with a shape of NCHW, the channel feature map refers to a 2-dimensional feature map with the shape of HW).

For example, the \(j_th\) channel of the \(i_th\) sample in the batched input is a 2D tensor input[i,j]. Each channel will be zeroed out independently on every forward call with probability 1-keep_prob using samples from a Bernoulli distribution.

Dropout2D can improve the independence between channel feature maps.

Parameters

keep_prob (float) – The keep probability of a channel, between 0 and 1, e.g. keep_prob = 0.8, means dropping out 20% of channels. Default: 0.5.

Inputs:
  • x (Tensor) - A 4-D tensor with shape \((N, C, H, W)\). The data type should be int8, int16, int32, int64, float16 or float32.

Outputs:
  • output (Tensor) - With the same shape and data type as x.

  • mask (Tensor) - With the same shape as x and the data type is bool.

Raises
  • TypeError – If the data type of keep_prob is not float.

  • ValueError – If keep_prob is out of the range [0.0, 1.0]; or if the dim of input is not 4-D.

Supported Platforms:

Ascend

Examples

>>> dropout = ops.Dropout2D(keep_prob=0.5)
>>> x = Tensor(np.ones([2, 1, 2, 3]), mindspore.float32)
>>> output, mask = dropout(x)
>>> print(output.shape)
(2, 1, 2, 3)