GeUtils
【免费下载链接】geGE（Graph Engine）是面向昇腾的图编译器和执行器，提供了计算图优化、多流并行、内存复用和模型下沉等技术手段，加速模型执行效率，减少模型内存占用。 GE 提供对 PyTorch、TensorFlow 前端的友好接入能力，并同时支持 onnx、pb 等主流模型格式的解析与编译。项目地址: https://gitcode.com/cann/ge
Product Support Status
| Product | Support Status |
|---|---|
| Atlas A3 Training Series Products/Atlas A3 Inference Series Products | √ |
| Atlas A2 Training Series Products/Atlas A2 Inference Series Products | √ |
Module Import
from ge.utils import GeUtilsFunctionality Description
GeUtilsprovides GE (Graph Engine) common utility interfaces, including Shape derivation and AICore operator support checking functionality.
infer_shape: Performs Shape derivation on computation graph. This interface only executes Shape derivation, does not perform other graph optimizations (such as constant folding, dead edge elimination etc.).check_node_support_on_aicore: Checks if node is supported on AICore, returns boolean value for support status and unsupported reason string.
Classes
GeUtils
GeUtilsclass only contains static methods, can be used without instantiation.
| Static Method | Description |
|---|---|
| infer_shape | Performs Shape derivation on computation graph |
| check_node_support_on_aicore | Checks if node is supported on AICore |
Function Prototypes
infer_shape
@staticmethod def infer_shape(graph: Graph, input_shapes: List[List[int]]) -> NonePerforms Shape derivation on computation graph. Only executes Shape derivation, does not perform constant folding, dead edge elimination and other graph optimizations.
check_node_support_on_aicore
@staticmethod def check_node_support_on_aicore(node: Node) -> Tuple[bool, str]Checks if node is supported on AICore, returns boolean value for support status and unsupported reason.
Parameter Description
infer_shape
| Parameter | Type | Required | Description |
|---|---|---|---|
| graph | Graph | Yes | Computation graph object to perform Shape derivation |
| input_shapes | List[List[int]] | Yes | Input shape list. Each element in list describes shape of a graph input, each shape is integer dimension list |
check_node_support_on_aicore
| Parameter | Type | Required | Description |
|---|---|---|---|
| node | Node | Yes | Computation graph node object to be checked |
Return Value Description
infer_shape
No return value. Shape derivation result is directly updated in graph object.
check_node_support_on_aicore
| Return Value | Type | Description |
|---|---|---|
| is_supported | bool | Whether node is supported on AICore. True means supported, False means not supported |
| unsupported_reason | str | Description of unsupported reason. If node is supported, returns empty string |
Constraint Description
infer_shapeonly executes Shape derivation, does not perform constant folding, dead edge elimination and other graph optimization operations.input_shapesforinfer_shapemust be list of lists, each sublist elements must be integer type.check_node_support_on_aicorerequires passing validNodeobject.- All methods in
GeUtilsare static methods, can be called without creating instance.
Usage Example
from ge.utils import GeUtils from ge.graph import Graph, Node # Shape derivation graph = Graph("my_graph") # ... build graph ... GeUtils.infer_shape(graph, [[1, 3, 224, 224], [1, 3, 224, 224]]) # Check if node is supported on AICore【免费下载链接】geGE（Graph Engine）是面向昇腾的图编译器和执行器，提供了计算图优化、多流并行、内存复用和模型下沉等技术手段，加速模型执行效率，减少模型内存占用。 GE 提供对 PyTorch、TensorFlow 前端的友好接入能力，并同时支持 onnx、pb 等主流模型格式的解析与编译。项目地址: https://gitcode.com/cann/ge
创作声明：本文部分内容由AI辅助生成（AIGC），仅供参考
