From 6eb1e63dd0ff6ad8c450264e60bb3713ad6a4fc4 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E5=90=95=E5=87=AF=E7=9B=9F?= Date: Fri, 21 Jun 2024 02:55:36 +0000 Subject: [PATCH 1/3] =?UTF-8?q?=E4=BF=AE=E6=94=B9=E5=8A=A8=E6=80=81?= =?UTF-8?q?=E5=9B=BE=E5=9C=BA=E6=99=AFdump=E6=96=87=E6=A1=A3?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: 吕凯盟 --- tutorials/experts/source_zh_cn/debug/dump.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tutorials/experts/source_zh_cn/debug/dump.md b/tutorials/experts/source_zh_cn/debug/dump.md index e2b84a8c7c..fb9f554bed 100644 --- a/tutorials/experts/source_zh_cn/debug/dump.md +++ b/tutorials/experts/source_zh_cn/debug/dump.md @@ -8,7 +8,7 @@ - 对于静态图模式，MindSpore提供了Dump功能，用来将模型训练中的图以及算子的输入输出数据保存到磁盘文件。 -- 对于动态图模式，Dump功能仅支持Ascend后端的溢出检测能力。要想查看非溢出节点，可以使用Python原生执行能力，用户可以在网络脚本运行过程中查看记录相应的输入输出。 +- 对于动态图模式，前向过程可以使用Python原生执行能力，用户可以在网络脚本运行过程中查看记录相应的输入输出。jit以及反向过程可以使用同步dump功能，将算子的输入输出数据保存到磁盘文件。 ### 调试过程 -- Gitee From e5e29046b7f26653aec9094bc4d37a8f82cb988f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E5=90=95=E5=87=AF=E7=9B=9F?= Date: Fri, 21 Jun 2024 06:30:10 +0000 Subject: [PATCH 2/3] =?UTF-8?q?=E4=BF=AE=E6=94=B9=E5=8A=A8=E6=80=81?= =?UTF-8?q?=E5=9B=BE=E5=9C=BA=E6=99=AFdump=E6=96=87=E6=A1=A3?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: 吕凯盟 --- tutorials/experts/source_en/debug/dump.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tutorials/experts/source_en/debug/dump.md b/tutorials/experts/source_en/debug/dump.md index e02393e3b3..e81afaa979 100644 --- a/tutorials/experts/source_en/debug/dump.md +++ b/tutorials/experts/source_en/debug/dump.md @@ -6,7 +6,7 @@ The input and output of the operator can be saved for debugging through the data dump when the training result deviates from the expectation. -- For the dynamic graph mode, the Dump function only support overflow detection ability on Ascend. To view those nodes which are not overflow, please use the native Python execution capabilities. Users can view and record the corresponding input and output during the running of the network script. +- For dynamic graph mode, the forward process can utilize Python's native execution capabilities, allowing users to view and record the corresponding inputs and outputs during the execution of the network script. The JIT and backward processes, which are part of graph compilation, can use synchronous dump functionality to save the input and output data of operators to disk files. - For the static graph mode, MindSpore provides the Dump function to save the graph and the input and output data of the operator during model training to a disk file. -- Gitee From eb0fe7402bf25023002d5ed9dd04e60868b87ab6 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E5=90=95=E5=87=AF=E7=9B=9F?= Date: Fri, 21 Jun 2024 06:32:15 +0000 Subject: [PATCH 3/3] =?UTF-8?q?=E4=BF=AE=E6=94=B9=E5=8A=A8=E6=80=81?= =?UTF-8?q?=E5=9B=BE=E5=9C=BA=E6=99=AFdump=E6=96=87=E6=A1=A3?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: 吕凯盟 --- tutorials/experts/source_zh_cn/debug/dump.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tutorials/experts/source_zh_cn/debug/dump.md b/tutorials/experts/source_zh_cn/debug/dump.md index fb9f554bed..e217a1a7a4 100644 --- a/tutorials/experts/source_zh_cn/debug/dump.md +++ b/tutorials/experts/source_zh_cn/debug/dump.md @@ -8,7 +8,7 @@ - 对于静态图模式，MindSpore提供了Dump功能，用来将模型训练中的图以及算子的输入输出数据保存到磁盘文件。 -- 对于动态图模式，前向过程可以使用Python原生执行能力，用户可以在网络脚本运行过程中查看记录相应的输入输出。jit以及反向过程可以使用同步dump功能，将算子的输入输出数据保存到磁盘文件。 +- 对于动态图模式，前向过程可以使用Python原生执行能力，用户可以在网络脚本运行过程中查看记录相应的输入输出。jit以及反向过程属于图编译的部分可以使用同步Dump功能，将算子的输入输出数据保存到磁盘文件。 ### 调试过程 -- Gitee