From da30cc0cc3b297a9f2c6b6b1fd2622f79dd4f93f Mon Sep 17 00:00:00 2001 From: xuzhen Date: Wed, 10 Dec 2025 10:34:57 +0800 Subject: [PATCH] modify table --- docs/mindspore/source_en/api_python/env_var_list.rst | 6 ++++-- docs/mindspore/source_zh_cn/api_python/env_var_list.rst | 6 ++++-- 2 files changed, 8 insertions(+), 4 deletions(-) diff --git a/docs/mindspore/source_en/api_python/env_var_list.rst b/docs/mindspore/source_en/api_python/env_var_list.rst index 8c90dcad20..3c2ef720a9 100644 --- a/docs/mindspore/source_en/api_python/env_var_list.rst +++ b/docs/mindspore/source_en/api_python/env_var_list.rst @@ -359,8 +359,10 @@ Graph Compilation and Execution - Control whether the Dispatch feature is enabled for operators in PyNative mode. When enabled, operators automatically selects the corresponding device for execution based on the device where the input tensor is located. - Integer - 1: Enable dispatch and disable implicit copying of operator inputs in PyNative mode. If the input tensors are on different devices, an exception will be thrown. - - 2: Enable dispatch in PyNative mode. If the input tensors are on different devices, the Python stack trace will be printed, and the operator will run on the higher-priority device (for example, if both Ascend and CPU tensors exist, Ascend will be preferred). - - 3: Enable dispatch in PyNative mode. If the input tensors are on different devices, the operator will run on the higher-priority device. + + 2: Enable dispatch in PyNative mode. If the input tensors are on different devices, the Python stack trace will be printed, and the operator will run on the higher-priority device (for example, if both Ascend and CPU tensors exist, Ascend will be preferred). + + 3: Enable dispatch in PyNative mode. If the input tensors are on different devices, the operator will run on the higher-priority device. Not set or any other value: Dispatch is disabled. - diff --git a/docs/mindspore/source_zh_cn/api_python/env_var_list.rst b/docs/mindspore/source_zh_cn/api_python/env_var_list.rst index 00d6e812f1..c999bc2291 100644 --- a/docs/mindspore/source_zh_cn/api_python/env_var_list.rst +++ b/docs/mindspore/source_zh_cn/api_python/env_var_list.rst @@ -359,8 +359,10 @@ - 控制在动态图流程中算子是否启用Dispatch功能。启用后，算子会根据输入tensor所在的设备，自动选择对应的设备执行。 - Integer - 1：开启动态图Dispatch功能，关闭动态图算子输入隐式拷贝。如果输入多个tensor的设备不一致，则会抛异常。 - - 2：开启动态图Dispatch功能。如果输入多个tensor的设备不一致，则会打印堆栈信息，然后按高优先级的设备执行（例如：同时存在Ascend和CPU的tensor，则优先到Ascend执行）。 - - 3：开启动态图Dispatch功能。如果输入多个tensor设备不一致，按高优先级的设备执行。 + + 2：开启动态图Dispatch功能。如果输入多个tensor的设备不一致，则会打印堆栈信息，然后按高优先级的设备执行（例如：同时存在Ascend和CPU的tensor，则优先到Ascend执行）。 + + 3：开启动态图Dispatch功能。如果输入多个tensor设备不一致，按高优先级的设备执行。 不设置或其他值：不开启Dispatch功能 - -- Gitee