Skip to content
体验新版
项目
组织
正在加载...
登录
切换导航
打开侧边栏
PaddlePaddle
Paddle
提交
faaa95ca
P
Paddle
项目概览
PaddlePaddle
/
Paddle
1 年多 前同步成功
通知
2302
Star
20931
Fork
5422
代码
文件
提交
分支
Tags
贡献者
分支图
Diff
Issue
1423
列表
看板
标记
里程碑
合并请求
543
Wiki
0
Wiki
分析
仓库
DevOps
项目成员
Pages
P
Paddle
项目概览
项目概览
详情
发布
仓库
仓库
文件
提交
分支
标签
贡献者
分支图
比较
Issue
1,423
Issue
1,423
列表
看板
标记
里程碑
合并请求
543
合并请求
543
Pages
分析
分析
仓库分析
DevOps
Wiki
0
Wiki
成员
成员
收起侧边栏
关闭侧边栏
动态
分支图
创建新Issue
提交
Issue看板
未验证
提交
faaa95ca
编写于
7月 04, 2022
作者:
Y
yaozhixin
提交者:
GitHub
7月 04, 2022
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
update paddle.distributed.launch en doc (#44016)
上级
91c0f727
变更
1
隐藏空白更改
内联
并排
Showing
1 changed file
with
9 addition
and
7 deletion
+9
-7
python/paddle/distributed/launch/main.py
python/paddle/distributed/launch/main.py
+9
-7
未找到文件。
python/paddle/distributed/launch/main.py
浏览文件 @
faaa95ca
...
...
@@ -98,11 +98,11 @@ def launch():
The ``training_script_args`` includes arguments required by IPU distributed launch and illustrated as below.
``Examples 10`` has provided a example of paddle.distributed.launch with IPUs.
- ``--hosts``: The hosts for IPU distributd training.
- ``--hosts``: The hosts for IPU distributd training.
Each host is able to include multiple processes.
- ``--nproc_per_host``: The number of processes launched per host.
- ``--nproc_per_host``: The number of processes launched per host.
Each process is able to include multiple replicas.
- ``--ipus_per_replica``: The number of IPUs requested per replica.
- ``--ipus_per_replica``: The number of IPUs requested per replica.
Each replica is able to include multiple IPUs.
- ``--ipu_partition``: The partition name of IPU devices.
...
...
@@ -110,7 +110,7 @@ def launch():
- ``training_script``: The full path to the IPU distributed training program/script to be launched in parallel. e.g., ``training.py``.
- ``training_script_args``: The args of the IPU distributed training program/script.
- ``training_script_args``: The args of the IPU distributed training program/script.
e.g., ``--lr=0.1``.
Returns:
- ``None``
...
...
@@ -253,9 +253,11 @@ def launch():
.. code-block:: bash
:name: code-block-example-bash10
# With the following command, the job will begin to run the distributhed program with IPUs.
# Only support and require the `device_num` as the arg and `ipu` as the launch script.
# Please Check the details about the following args of the launch scripte from `utils/ipu_launch.py`.
# With the following command, the job will begin to run the distributhed program with IPUs
# Require `devices` as the number of IPUs
# Require `training_script` to be set as `ipu`
# Require `training_script_args` as the arguments of IPU distributed training instead of the arguments of the training program/script
# Please Check the `IPU Parameters` for details
python -m paddle.distributed.launch --devices 4 ipu --hosts=localhost --nproc_per_host=2 --ipus_per_replica=1 --ipu_partition=pod16 --vipu_server=127.0.0.1 train.py
"""
...
...
编辑
预览
Markdown
is supported
0%
请重试
或
添加新附件
.
添加附件
取消
You are about to add
0
people
to the discussion. Proceed with caution.
先完成此消息的编辑!
取消
想要评论请
注册
或
登录