FilesLast commitLast update
6 days ago
6 days ago
3 months ago
6 days ago
6 days ago
6 days ago
6 days ago
6 days ago
6 days ago
1 month ago
6 days ago
6 days ago
README

Asinh

English | 简体中文

The inverse hyperbolic sine (Inverse Hyperbolic Sine) operator, corresponding to PyTorch torch.asinh / torch.asinh_. Online aclnn API documentation: docs/aclnnAsinh&aclnnInplaceAsinh_en.md

Product Support Status

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

Function Description

  • Interface function: performs the inverse hyperbolic sine operation on each element in the input Tensor and outputs the result.

  • Computation formula:

yi=ln(xi+xi2+1)y_{i}=ln(x_{i} + \sqrt{x_{i}^2 + 1})

Parameter Description

Parameter Name Input/Output/Attribute Description Data Type Data Format
x Input Input x in the formula, whose dtype is the same as that of y. The shape is arbitrary (1 to 8 dimensions). Empty Tensors are supported, and non-contiguous Tensors are supported. FLOAT, FLOAT16, BFLOAT16 ND
y Output Output y in the formula. The shape is exactly the same as that of x. It is specific to the non-in-place API. FLOAT, FLOAT16, BFLOAT16 ND
selfRef Input/Output Specific to the in-place API. It is a tensor that serves as both input and output, and the computation result is written back in place. FLOAT, FLOAT16, BFLOAT16 ND
  • Ascend 950PR&950DT products, Atlas A2 products, and Atlas A3 products: support BFLOAT16.
  • Atlas training products and Atlas inference products: do not support BFLOAT16.
  • Input types additionally supported by the aclnn interface: INT8, INT16, INT32, INT64, UINT8, and BOOL are converted to FLOAT32 for computation; DOUBLE, COMPLEX64, and COMPLEX128 are computed through the AICPU path.

Constraint Description

None

Invocation Description

Invocation Method Invocation Sample Description
aclnn invocation (non-in-place) examples/test_aclnn_asinh.cpp Refer to build.sh in the root directory of the ops-math repository to complete operator compilation and operator package installation, and then verify with the examples (on a physical NPU).
Graph-mode (GE IR) invocation examples/test_geir_asinh.cpp