FilesLast commitLast update
13 days ago
10 months ago
18 days ago
1 month ago
6 days ago
2 months ago
1 month ago
6 days ago
1 month ago
6 days ago
README

Fill

English | 简体中文

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

  • Operator function: performs a fill operation on tensors and supports non-contiguous Tensor operations.
  • InplaceFillScalar: fills the entire tensor with a scalar value.
  • InplaceFillTensor: fills the entire tensor with a tensor value (the value tensor must be a 0-dimensional tensor or a 1-dimensional tensor with size=1).

Parameter Description

InplaceFillScalar

Parameter Name Input/Output/Attribute Description Data Type Data Format
selfRef Input/Output tensor Input and output tensor. Non-contiguous Tensors are supported, and data dimensions above 8 are not supported. FLOAT, FLOAT16, UINT8, INT8, INT16, INT32, INT64, DOUBLE, COMPLEX64, COMPLEX128, BOOL, BFLOAT16, UINT16, UINT32, UINT64, FLOAT8_E4M3FN, FLOAT8_E5M2, FLOAT8_E8M0, HIFLOAT8 ND
value Input scalar Specified scalar value. Its data type must satisfy the data type inference rules with the data type of selfRef. FLOAT, FLOAT16, UINT8, INT8, INT16, INT32, INT64, DOUBLE, COMPLEX64, COMPLEX128, BOOL, BFLOAT16, UINT16, UINT32, UINT64, FLOAT8_E4M3FN, FLOAT8_E5M2, FLOAT8_E8M0, HIFLOAT8 -

InplaceFillTensor

Parameter Name Input/Output/Attribute Description Data Type Data Format
selfRef Input/Output tensor Input and output tensor. Non-contiguous Tensors are supported, and data dimensions above 8 are not supported. FLOAT, FLOAT16, UINT8, INT8, INT16, INT32, INT64, DOUBLE, COMPLEX64, COMPLEX128, BOOL, BFLOAT16, FLOAT8_E4M3FN, FLOAT8_E5M2, FLOAT8_E8M0, HIFLOAT8 ND
value Input tensor Fill value tensor. It must be a 0-dimensional tensor or a 1-dimensional tensor with size=1. FLOAT, FLOAT16, UINT8, INT8, INT16, INT32, INT64, DOUBLE, COMPLEX64, COMPLEX128, BOOL, BFLOAT16, FLOAT8_E4M3FN, FLOAT8_E5M2, FLOAT8_E8M0, HIFLOAT8 ND

Constraint Description

  • Data dimensions above 8 are not supported for the input tensor.
  • The value parameter of InplaceFillTensor must be a 0-dimensional tensor or a 1-dimensional tensor with size=1.
  • Data types must satisfy the mutual inference relationship.
  • The FLOAT8_E4M3FN, FLOAT8_E5M2, FLOAT8_E8M0, and HIFLOAT8 data types are supported only on Ascend 950PR&950DT products and are not supported on other products.
  • For CANN 9.0.0, the FLOAT8_E4M3FN, FLOAT8_E5M2, FLOAT8_E8M0, and HIFLOAT8 data types are not supported.

Invocation Description

Invocation Method Sample Code Description
aclnn API test_aclnn_fill_scalar.cpp Invokes the InplaceFillScalar operator through the aclnnInplaceFillScalar API.
aclnn API test_aclnn_fill_tensor.cpp Invokes the InplaceFillTensor operator through the aclnnInplaceFillTensor API.