[PowerSGD] Add orthogonalization with QR factorization #72043
Closed
Conversation
This file contains 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
younik
requested review from
H-Huang,
mingzhe09088,
mrshenli,
pritamdamania87,
rohan-varma and
zhaojuanmao
as code owners
CI Flow Status⚛️ CI FlowRuleset - Version:
|
🔗 Helpful links
💊 CI failures summary and remediationsAs of commit 488124f (more details on the Dr. CI page):
🕵️ 1 new failure recognized by patternsThe following CI failures do not appear to be due to upstream breakages:
|
| Job | Step | Action |
|---|---|---|
 linux-bionic-rocm4.5-py3.7 / test (default, 1, 2, linux.rocm.gpu) |
Test | 🔁 rerun |
This comment was automatically generated by Dr. CI (expand for details).
Please report bugs/suggestions to the (internal) Dr. CI Users group.
facebook-github-bot
added
the
oncall: distributed
albanD
added
the
triaged
|
Overall looks good to me. One question though; as we discussed offline previously, Thijs experiments showed performance improvement with a rank of 2 as well. Do you know the reason why you couldn't observe the same effect in your experiments? |
Great question; I get results similar to Thijs when monitoring only the _orthogonalize method. On the other hand, I obtain the above results monitoring the whole powerSGD_hook. I am not sure why this is happening. |
|
There should be now |
cbalioglu
approved these changes
Feb 7, 2022
|
@cbalioglu has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator. |
facebook-github-bot
pushed a commit
that referenced
this pull request
Feb 7, 2022
Summary: ### 🚀 The feature, motivation and pitch Following the discussion in #65813, I added the QR factorization to powerSGD_hook.py Gram-Schmidt orthogonalization can't be fully replaced because _torch.linalg.qr_ doesn't work with half-precision. Moreover, in my tests, it works faster with a rank lesser than 3. This is one sample experiment timing powerSGD_hook on ResNext101 with the two different methods:  ### Alternatives Use _torch.orgqr(*torch.geqrf(matrix))_. From my tests it performances are similar to _torch.linalg.qr_. ### Additional context _No response_ Pull Request resolved: #72043 Reviewed By: albanD Differential Revision: D34042781 Pulled By: cbalioglu fbshipit-source-id: e331179d3b7ac40d445b651fc473b16ae4ead462
|
Hey @younik. |
cbalioglu
added
release notes: distributed (ddp)
cyyever
pushed a commit
to cyyever/pytorch_private
that referenced
this pull request
Feb 9, 2022
Summary: ### 🚀 The feature, motivation and pitch Following the discussion in pytorch/pytorch#65813, I added the QR factorization to powerSGD_hook.py Gram-Schmidt orthogonalization can't be fully replaced because _torch.linalg.qr_ doesn't work with half-precision. Moreover, in my tests, it works faster with a rank lesser than 3. This is one sample experiment timing powerSGD_hook on ResNext101 with the two different methods:  ### Alternatives Use _torch.orgqr(*torch.geqrf(matrix))_. From my tests it performances are similar to _torch.linalg.qr_. ### Additional context _No response_ Pull Request resolved: pytorch/pytorch#72043 Reviewed By: albanD Differential Revision: D34042781 Pulled By: cbalioglu fbshipit-source-id: e331179d3b7ac40d445b651fc473b16ae4ead462 (cherry picked from commit f64bf3839aad795fc0ad12da15fa2e9a0decf5ab)
cyyever
pushed a commit
to cyyever/pytorch_private
that referenced
this pull request
Feb 9, 2022
Summary: ### 🚀 The feature, motivation and pitch Following the discussion in pytorch/pytorch#65813, I added the QR factorization to powerSGD_hook.py Gram-Schmidt orthogonalization can't be fully replaced because _torch.linalg.qr_ doesn't work with half-precision. Moreover, in my tests, it works faster with a rank lesser than 3. This is one sample experiment timing powerSGD_hook on ResNext101 with the two different methods:  ### Alternatives Use _torch.orgqr(*torch.geqrf(matrix))_. From my tests it performances are similar to _torch.linalg.qr_. ### Additional context _No response_ Pull Request resolved: pytorch/pytorch#72043 Reviewed By: albanD Differential Revision: D34042781 Pulled By: cbalioglu fbshipit-source-id: e331179d3b7ac40d445b651fc473b16ae4ead462 (cherry picked from commit f64bf3839aad795fc0ad12da15fa2e9a0decf5ab)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
🚀 The feature, motivation and pitch
Following the discussion in #65813, I added the QR factorization to powerSGD_hook.py
Gram-Schmidt orthogonalization can't be fully replaced because torch.linalg.qr doesn't work with half-precision. Moreover, in my tests, it works faster with a rank lesser than 3.
This is one sample experiment timing powerSGD_hook on ResNext101 with the two different methods:
Alternatives
Use torch.orgqr(*torch.geqrf(matrix)). From my tests it performances are similar to torch.linalg.qr.
Additional context
No response