translate : recipes_source/distributed_device_mesh.rst#1132
Merged
jih0-kim merged 4 commits intoJun 1, 2026
Conversation
Contributor
|
์๋ ํ์ธ์ ๋์๋, ์ ์ฒด์ ์ผ๋ก ๊น๋ํ ๋ฒ์ญ์ธ๊ฑฐ ๊ฐ์ต๋๋ค, LGTM! |
Contributor
|
line 177์์ ์ฐธ๊ณ ๋ฌธ์ ์ ๋ชฉ์ด ๋ฒ์ญ๋์ด ์๋๋ฐ, line 178์ฒ๋ผ ์๋ฌธ ๊ทธ๋๋ก ๋๋ ๊ฒ ๋ซ์ง ์์๊น ์ถ์ต๋๋ค. |
hyoyoung
requested changes
May 28, 2026
Member
hyoyoung
left a comment
hyoyoung
left a comment
There was a problem hiding this comment.
๊ธด ๋ฌธ์ ๋ฒ์ญํ๋๋ผ ์๊ณ ํ์
จ์ต๋๋ค.
์ฌ์ํ ์ ์ ์ฌํญ ๋ช๊ฐ์ง ํ์ธ๋ถํ๋๋ฆฝ๋๋ค
| ===================================================== | ||
|
|
||
| **Author**: `Iris Zhang <https://github.com/wz337>`__, `Wanchao Liang <https://github.com/wanchaol>`__ | ||
| **์ ์**: `Iris Zhang <https://github.com/wz337>`__, `Wanchao Liang <https://github.com/wanchaol>`__ |
Member
There was a problem hiding this comment.
์ ์ ์๋์ ์ญ์ ํญ๋ชฉ ์ถ๊ฐ ๋ถํ๋๋ฆฝ๋๋ค
| DeviceMesh is useful when working with multi-dimensional parallelism (i.e. 3-D parallel) where parallelism composability is required. For example, when your parallelism solutions require both communication across hosts and within each host. | ||
| The image above shows that we can create a 2D mesh that connects the devices within each host, and connects each device with its counterpart on the other hosts in a homogeneous setup. | ||
| DeviceMesh๋ ์ฌ๋ฌ ๋ณ๋ ฌํ ๋ฐฉ์์ ์กฐํฉ(composability)ํด์ผ ํ๋ ๋ค์ฐจ์ ๋ณ๋ ฌํ(์: 3-D ๋ณ๋ ฌ)๋ฅผ ๋ค๋ฃฐ ๋ ์ ์ฉํฉ๋๋ค. ์๋ฅผ ๋ค์ด, ๋ณ๋ ฌํ ๋ฐฉ์์ด ํธ์คํธ ๊ฐ ํต์ ๊ณผ ๊ฐ ํธ์คํธ ๋ด๋ถ์ ํต์ ์ ๋ชจ๋ ์๊ตฌํ๋ ๊ฒฝ์ฐ๊ฐ ๊ทธ๋ ์ต๋๋ค. | ||
| ์ ์ด๋ฏธ์ง๋ ๊ท ์ผํ ํ๊ฒฝ์์ ๊ฐ ํธ์คํธ ๋ด๋ถ์ ๋๋ฐ์ด์ค๋ฅผ ์ฐ๊ฒฐํ๊ณ , ๊ฐ ๋๋ฐ์ด์ค๋ฅผ ๋ค๋ฅธ ํธ์คํธ์ ๋์ ๋๋ฐ์ด์ค์ ์ฐ๊ฒฐํ๋ 2D ๋ฉ์๋ฅผ ๋ง๋ค ์ ์์์ ๋ณด์ฌ์ค๋๋ค. |
Member
There was a problem hiding this comment.
๊ท ์ผํ ํ๊ฒฝ๋ ์ข์ง๋ง ๋์ผํ ๊ตฌ์ฑ์ ํ๊ฒฝ์ผ๋ก ๋ฐ๊พธ๋ฉด ์์ฐ์ค๋ฌ์์ง๊ฑฐ ๊ฐ์ต๋๋ค
| First, we need to manually calculate the shard group and replicate group. Then, we need to assign the correct shard and | ||
| replicate group to each rank. | ||
| DeviceMesh๊ฐ ์๋ค๋ฉด, ์ด๋ค ๋ณ๋ ฌํ๋ฅผ ์ ์ฉํ๊ธฐ ์ ์ ๊ฐ ํ๋ก์ธ์ค๋ง๋ค NCCL ํต์ ๊ธฐ์ CUDA ๋๋ฐ์ด์ค๋ฅผ ์ง์ ์ค์ ํด์ผ ํ๋ฉฐ, ์ด๋ ๊ฝค ๋ณต์กํ ์์ ์ ๋๋ค. | ||
| ๋ค์ ์ฝ๋๋ :class:`DeviceMesh` ์์ด ํ์ด๋ธ๋ฆฌ๋ ์ค๋ฉ(hybrid sharding) 2-D ๋ณ๋ ฌ ํจํด์ ์ค์ ํ๋ ์์์ ๋๋ค. |
Member
There was a problem hiding this comment.
2-D๋ณด๋ค๋ 2์ฐจ์ ์ ๋๋ก ๋ฐ๊พธ๋๊ฒ์ ์ด๋จ๊น์
| With the help of :func:`init_device_mesh`, we can accomplish the above 2D setup in just two lines, and we can still | ||
| access the underlying :class:`ProcessGroup` if needed. | ||
| :func:`init_device_mesh` ๋ฅผ ํ์ฉํ๋ฉด ์์ 2D ์ค์ ์ ๋จ ๋ ์ค๋ก ๋๋ผ ์ ์๊ณ , ํ์ํ ๋๋ | ||
| ๋ด๋ถ์ :class:`ProcessGroup` ์๋ ๊ทธ๋๋ก ์ ๊ทผํ ์ ์์ต๋๋ค. |
Member
There was a problem hiding this comment.
๊ทธ๋๋ก๋ ์์ด๋ ๋ ๊ฑฐ ๊ฐ์ต๋๋ค
| -------------------------------------------------------- | ||
| When working with large scale training, you might have more complex custom parallel training composition. For example, you may need to slice out sub-meshes for different parallelism solutions. | ||
| DeviceMesh allows users to slice child mesh from the parent mesh and re-use the NCCL communicators already created when the parent mesh is initialized. | ||
| ๋๊ท๋ชจ ํ์ต ํ๊ฒฝ์์๋ ๋ ๋ณต์กํ ์ฌ์ฉ์ ์ ์ ๋ณ๋ ฌ ํ์ต ๊ตฌ์ฑ์ ๋ค๋ค์ผ ํ ์๋ ์์ต๋๋ค. ์๋ฅผ ๋ค์ด, ์๋ก ๋ค๋ฅธ ๋ณ๋ ฌํ ๋ฐฉ์์ ๋ง์ถฐ ํ์ ๋ฉ์(sub-mesh)๋ฅผ ์๋ผ๋ด์ผ ํ ์ ์์ต๋๋ค. |
Member
There was a problem hiding this comment.
์๋ผ๋ด๋๊ฒ ์๋ฏธ๋ ๋ง๋๋ฐ ์กฐ๊ธ ๋ ์์ญํด๋ ์ข์๊ฑฐ ๊ฐ์ต๋๋ค
ํ์ ๋ฉ์๋ฅผ ๋๋์ด ์ฌ์ฉํด์ผ ํ ์ ์์ต๋๋ค ์ ๋๋ ์ด๋จ๊น์
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
๋ผ์ด์ ์ค ๋์
๋ณ๊ฒฝํด์ฃผ์๋ ๋ด์ฉ์ BSD 3ํญ ๋ผ์ด์ ์ค๊ฐ ์ ์ฉ๋จ์ ๋์ํด์ฃผ์ ์ผ ํฉ๋๋ค.
๋ ์์ธํ ๋ด์ฉ์ ๊ธฐ์ฌํ๊ธฐ ๋ฌธ์๋ฅผ ์ฐธ๊ณ ํด์ฃผ์ธ์.
๋์ํ์๋ฉด ์๋
[ ]๋ฅผ[x]๋ก ๋ง๋ค์ด์ฃผ์ธ์.๊ด๋ จ ์ด์ ๋ฒํธ
์ด Pull Request์ ๊ด๋ จ์๋ ์ด์ ๋ฒํธ๋ฅผ ์ ์ด์ฃผ์ธ์.
์ด์ ๋๋ PR ๋ฒํธ ์์ #์ ๋ถ์ด์๋ฉด ์ ๋ชฉ์ ๋ฐ๋ก ํ์ธํ์ค ์ ์์ต๋๋ค. (์. #999 )
PR ์ข ๋ฅ
์ด PR์ ํด๋น๋๋ ์ข ๋ฅ ์์
[ ]์[x]๋ก ๋ณ๊ฒฝํด์ฃผ์ธ์.PR ์ค๋ช
recipes_source/distributed_device_mesh.rst ๋ฌธ์๋ฅผ ๋ฒ์ญํ์์ต๋๋ค.