mindspore.amp.get_white_list

View Source On Gitee
mindspore.amp.get_white_list()[source]

Provide a copy of internal white list used by auto mixed precision with amp_level set to O1.

The current built-in whitelist contents are:

[mindspore.nn.Conv1d, mindspore.nn.Conv2d, mindspore.nn.Conv3d, mindspore.nn.Conv1dTranspose, mindspore.nn.Conv2dTranspose, mindspore.nn.Conv3dTranspose, mindspore.nn.Dense, mindspore.nn.LSTMCell, mindspore.nn.RNNCell, mindspore.nn.GRUCell, mindspore.ops.Conv2D, mindspore.ops.Conv3D, mindspore.ops.Conv2DTranspose, mindspore.ops.Conv3DTranspose, mindspore.ops.MatMul, mindspore.ops.BatchMatMul, mindspore.ops.PReLU, mindspore.ops.ReLU, mindspore.ops.Ger]

Returns

list, A copy of internal white list.

Examples

>>> from mindspore import amp
>>> white_list = amp.get_white_list()
>>> print(white_list)
[<class 'mindspore.nn.layer.conv.Conv1d'>, <class 'mindspore.nn.layer.conv.Conv2d'>,
 <class 'mindspore.nn.layer.conv.Conv3d'>, <class 'mindspore.nn.layer.conv.Conv1dTranspose'>,
 <class 'mindspore.nn.layer.conv.Conv2dTranspose'>, <class 'mindspore.nn.layer.conv.Conv3dTranspose'>,
 <class 'mindspore.nn.layer.basic.Dense'>, <class 'mindspore.nn.layer.rnn_cells.LSTMCell'>,
 <class 'mindspore.nn.layer.rnn_cells.RNNCell'>, <class 'mindspore.nn.layer.rnn_cells.GRUCell'>,
 <class 'mindspore.ops.operations.nn_ops.Conv2D'>, <class 'mindspore.ops.operations.nn_ops.Conv3D'>,
 <class 'mindspore.ops.operations.nn_ops.Conv2DTranspose'>,
 <class 'mindspore.ops.operations.nn_ops.Conv3DTranspose'>,
 <class 'mindspore.ops.operations.nn_ops.Conv2DBackpropInput'>,
 <class 'mindspore.ops.operations.math_ops.MatMul'>, <class 'mindspore.ops.operations.math_ops.BatchMatMul'>,
 <class 'mindspore.ops.operations.nn_ops.PReLU'>, <class 'mindspore.ops.operations.nn_ops.ReLU'>,
 <class 'mindspore.ops.operations.math_ops.Ger'>]