FilesLast commitLast update
11 days ago
4 months ago
2 months ago
2 months ago
18 days ago
4 days ago
2 months ago
18 days ago
4 days ago
1 month ago
18 days ago
README

Mish

English | 简体中文

Applicable Products

Product Support
Ascend 950PR&950DT products √
Atlas A3 products √
Atlas A2 products √
Atlas 200I/500 A2 inference products ×
Atlas inference products ×
Atlas training products √

Function

  • Operator function: A self-regularized non-monotonic neural network activation function.

  • Formula:

outi=selfi×tanh⁡(Softplus(selfi))out_i = self_i \times \tanh(\mathrm{Softplus}(\mathrm{self}_i))

where Softplus(selfi)=ln⁡(1+exp⁡(selfi))\mathrm{Softplus}(self_i) = \ln{(1 + \exp{(self_i)})}

Parameters

Parameter name Input/Output/Attribute Description Data type Data format
self Input Input parameter for the Mish activation computation, corresponding to self in the calculation formula. FLOAT, FLOAT16, BFLOAT16 ND
out Output Output parameter of the Mish activation computation, corresponding to out in the calculation formula. FLOAT, FLOAT16, BFLOAT16 ND
  • Atlas training products: The supported data types are FLOAT and FLOAT16.

Constraints

  • Input shape restriction: The input dimensions cannot exceed 8, and only 0 to 8 dimensions are supported.

Invocation Description

Invocation method Invocation sample Description
aclnn invocation test_aclnn_mish Calls the Mish operator through the aclnnMish&aclnnInplaceMish interface.
aclnn invocation test_aclnn_inplace_mish Calls the Mish operator through the aclnnMish&aclnnInplaceMish interface.
Graph mode - Calls the Mish operator by building a graph through operator IR.