CANN/ge Shape类API文档

Shape

【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.graph import Shape

Functionality Description

Shape class inherits from Python built-inlist, used to represent tensor shape dimension information. Besides having all operation capabilities of standard lists, it also provides convenient methods for calculating total element count of shape and judging if it is unknown shape. When dims is None, represents scalar (empty list).

Shape module also defines the following constants:

| Constant Name | Value | Description | | | :----- | :--- | :--- | | UNKNOWN_DIM | -1 | Represents unknown dimension | | UNKNOWN_DIM_NUM | -2 | Represents unknown dimension count | | UNKNOWN_DIM_SIZE | -1 | Return value of get_shape_size() when shape is unknown |

Class Definition

class Shape(list): def __init__(self, dims: Optional[List[int]] = None) -> None

Function List

| Function | Functionality Description | | | :--- | :--- | | __init__(dims=None) | Constructor, creates Shape object. dims is integer list, None represents scalar (empty list) | | get_shape_size() | Calculates product of all dimensions in shape, i.e., total element count of tensor | | is_unknown_shape() | Judges if shape contains unknown dimensions |

Parameter Description

__init__ Parameter

| Parameter | Type | Required | Description | | | :----- | :--- | :------: | :--- | | dims | List[int] | No | Dimension value list, e.g., [1, 3, 224, 224]. None represents scalar (empty list). Default value is None |

Return Value Description

| Function | Return Type | Description | | | :--- | :--- | :--- | | get_shape_size() | int | Product of all dimensions. Returns 0 when shape is empty (scalar); returns -1 when shape contains unknown dimension (UNKNOWN_DIM or UNKNOWN_DIM_NUM) | | is_unknown_shape() | bool | Returns True if shape contains UNKNOWN_DIM (-1) or UNKNOWN_DIM_NUM (-2); otherwise returns False |

Constraint Description

• dims parameter must be integer list (list of int) or None, otherwise throws TypeError.
• Shape inherits from list, therefore supports all standard list operations (indexing, slicing, iteration, len etc.).
• When shape contains unknown dimension, get_shape_size() returns -1, rather than throwing exception.

Usage Example

from ge.graph import Shape # Create Shape object shape = Shape([1, 3, 224, 224]) # Get total element count print(shape.get_shape_size()) # 150528 # Judge if it is unknown shape print(shape.is_unknown_shape()) # False # Create Shape containing unknown dimension unknown_shape = Shape([-1, 3, 224, 224]) print(unknown_shape.is_unknown_shape()) # True print(unknown_shape.get_shape_size()) # -1 # Scalar shape (empty list) scalar = Shape() print(len(scalar)) # 0 print(scalar.get_shape_size()) # 0 # Supports list operations print(shape[0]) # 1 print(len(shape)) # 4 print(list(shape)) # [1, 3, 224, 224]

【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