FilesLast commitLast update
18 days ago
18 days ago
18 days ago
18 days ago
18 days ago
18 days ago
9 days ago
README

PartitionedCall

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: passes the input tensors args to the subgraph f and returns the output tensors of the subgraph f. The subgraph f receives the inputs and returns another group of tensors.

Parameter Description

Parameter Name Input/Output/Attribute Description Data Type Data Format
args Input (dynamic) Input tensors, which will be passed to the subgraph f and support dynamic inputs. All types ND
f Subgraph Subgraph that receives args and returns another group of tensors. - -
config Attribute Optional string, which is empty by default. String -
config_proto Attribute Optional string, which is empty by default. String -
executor_type Attribute Optional string, which is empty by default. String -
output Output (dynamic) Output tensors returned by the subgraph f, which support dynamic outputs. All types ND

Constraint Description

  • The number of args tensors must be the same as the number of output tensors.
  • The data type of each args tensor must be consistent with that of the corresponding output.

Invocation Description

Invocation Method Invocation Sample Description
Graph-mode invocation test_geir_partitioned_call Invokes the PartitionedCall operator by building a graph through the operator IR.