megengine.functional.nn.dropout

dropout(inp, drop_prob, training=True)

Returns a new tensor where each of the elements are randomly set to zero with probability P = drop_prob. Optionally rescale the output tensor if training is True.

Parameters

• inp (Tensor) – input tensor.

• drop_prob (float) – probability to drop (set to zero) a single element.

• training (bool) – the default behavior of dropout during training is to rescale the output, then it can be replaced by an Identity during inference. Default: True

Return type

Tensor

Returns

the ouput tensor

Examples

>>> import numpy as np
>>> data = Tensor(np.ones(10000000, dtype=np.float32))
>>> out = F.nn.dropout(data, 1.0 / 3.0, training=True)
>>> assert not out.numpy().all()
>>> out = F.nn.dropout(data, 1.0 / 3.0, training=False)
>>> assert out.numpy().all()
>>> out.numpy()
array([1., 1., 1., ..., 1., 1., 1.], dtype=float32)

previous

megengine.functional.nn.multi_head_attention

next

megengine.functional.nn.one_hot