--optypelist_for_implmode

Description

Sets the implementation mode of an operator on the optype list. The operator implementation mode can be either high_precision or high_performance.

See Also

• Use this option in conjunction with --op_select_implmode. Operators specified by --optypelist_for_implmode use the implementation mode specified by --op_select_implmode. It is not allowed to set different implementation modes for operators on the list.
• Use this option in conjunction with --op_select_implmode. This option is mutually exclusive with --op_precision_mode, which takes precedence over this option if they are both included. For details about the running process of the preceding parameters, see Figure 1.

Argument

Argument: Operator list.

Restrictions:

• The type of the operators on the list must be OpType of the Ascend IR–defined operator. For details about how to view the operator type, see How Do I Determine the Mapping Between Operators in the Original Network Model and Operators Supported by Ascend AI Processors?.
• The operators on the list use the implementation mode specified by --op_select_implmode, which is either high_precision or high_performance. Use commas (,) to separate operators.
• This argument takes effect only for specified operators. For others, the default implementation mode is used.

Suggestions and Benefits

None

Example

--op_select_implmode=high_precision  --optypelist_for_implmode=Pooling,SoftmaxV2

The preceding configuration example indicates that the high-precision mode is used only for the Pooling and SoftmaxV2 operators. For operators whose precision modes are not specified, the default implementation mode is used.

Applicability

Atlas 200/300/500 Inference Product

Atlas Training Series Product

Dependencies and Restrictions

None