mindchemistry.e3.nn.Activation

View Source On Gitee
class mindchemistry.e3.nn.Activation(irreps_in, acts, dtype=float32)[source]

Activation function for scalar-tensors. The parities of irreps may be changed according to the parity of each activation functions. Odd scalars require the corresponding activation functions to be odd or even.

Parameters

  • irreps_in (Union[str, Irrep, Irreps]) – the input irreps.

  • acts (List[Func]) – a list of activation functions for each part of irreps_in. The length of the acts will be clipped or filled by identity functions to match the length of irreps_in.

  • dtype (mindspore.dtype) – The type of input tensor. Default: mindspore.float32.

Inputs:

  • inputs (Tensor) - The shape of Tensor is \((*, irreps\_in.dim)\).

Outputs:

  • outputs (Tensor) - The shape of Tensor is \((*, irreps\_in.dim)\).

Raises

  • ValueError – If irreps_in contain non-scalar irrep.

  • ValueError – If a irrep in irreps_in is odd, but the corresponding activation function is neither even nor odd.

Supported Platforms:

Ascend

Examples

>>> from mindchemistry.e3.nn import Activation
>>> from mindspore import ops, Tensor
>>> act = Activation('3x0o+2x0e+1x0o', [ops.abs, ops.tanh])
>>> print(act)
Activation [xx-] (3x0o+2x0e+1x0o -> 3x0e+2x0e+1x0o)
>>> inputs = Tensor(ops.ones((4,6)))
>>> outputs = act(inputs)
>>> print(outputs.shape)
(4, 6)