Nigel
/
mindspore
forked from mindspore-Ecosystem/mindspore
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

!44327 Add meory offload doc 

Merge pull request !44327 from tanghuikang/code_docs_memory_offload
This commit is contained in:
i-robot 2022-10-27 08:16:24 +00:00 committed by Gitee
parent e720b5e117 6ef8965c32
commit a030fc8e1a
No known key found for this signature in database
GPG Key ID: 173E9B9CA92EEF8F
2 changed files with 17 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
docs/api/api_python/mindspore/mindspore.set_context.rst
View File

@ -74,6 +74,8 @@ mindspore.set_context
| | support_binary | CPU/GPU/Ascend |
| +------------------------------+----------------------------+
| | memory_optimize_level | CPU/GPU/Ascend |
| +------------------------------+----------------------------+
| | memory_offload | GPU/Ascend |
+-------------------------+------------------------------+----------------------------+
参数：
@ -146,6 +148,10 @@ mindspore.set_context
- O0: 执行性能优先，关闭 SOMAS (Safe Optimized Memory Allocation Solver)。
- O1: 内存性能优先，使能 SOMAS。
- **memory_offload** (str) - 是否开启Offload功能，在内存不足场景下将空闲数据临时拷贝至Host侧内存。其值必须在['ON', 'OFF']范围中，默认值为'OFF'。
- ON：开启memory offload功能。在Ascend硬件平台，未设置环境变量“GRAPH_OP_RUN=1”时本参数不生效；设置memory_optimize_level='O1'时本参数不生效。
- OFF：关闭memory offload功能。
异常：
- **ValueError** - 输入key不是上下文中的属性。

11
mindspore/python/mindspore/context.py
View File

@ -789,6 +789,8 @@ def set_context(**kwargs):
| | support_binary | CPU/GPU/Ascend |
| +------------------------------+----------------------------+
| | memory_optimize_level | CPU/GPU/Ascend |
| +------------------------------+----------------------------+
| | memory_offload | GPU/Ascend |
+-------------------------+------------------------------+----------------------------+
Args:
@ -938,6 +940,14 @@ def set_context(**kwargs):
- O0: priority performance option, disable SOMAS (Safe Optimized Memory Allocation Solver).
- O1: priority memory option, enable SOMAS.
memory_offload (str): Whether to enable the memory offload function. When it is enabled, the idle data will be
temporarily copied to the host side in the case of insufficient device memory. The value must be in the
range of ['ON', 'OFF'], and the default value is 'OFF'.
- ON: Enable the memory Offload function. On Ascend hardware platform, this parameter does not take effect
when the environment variable "GRAPH_OP_RUN=1" is not set; This parameter does not take effect when
memory_optimize_level is set 'O1'.
- OFF: Turn off the memory Offload function.
Raises:
ValueError: If input key is not an attribute in context.
@ -968,6 +978,7 @@ def set_context(**kwargs):
>>> ms.set_context(inter_op_parallel_num=4)
>>> ms.set_context(disable_format_transform=True)
>>> ms.set_context(memory_optimize_level='O0')
>>> ms.set_context(memory_offload='ON')
"""
ctx = _context()
# set device target first