Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Megatron Encoder Decoder models with RPE and PP > 2 #4663

Merged
merged 42 commits into from
Aug 26, 2022
Merged

Conversation

MaximumEntropy
Copy link
Contributor

Signed-off-by: MaximumEntropy [email protected]

What does this PR do ?

Adds support for Enc-Dec models with RPE and PP > 2

Collection: NLP

Changelog

  • Add specific line by line info of high level changes in this PR.

Usage

  • model.encoder.position_embedding_type="relative"
  • model.decoder.position_embedding_type="relative"
  • model.pipeline_model_parallel_size=4
  • model.pipeline_model_parallel_split_rank=2

Before your PR is "Ready for review"

Pre checks:

  • Make sure you read and followed Contributor guidelines
  • Did you write any new necessary tests?
  • Did you add or update any necessary documentation?
  • Does the PR affect components that are optional to install? (Ex: Numba, Pynini, Apex etc)
    • Reviewer: Does the PR have correct import guards for all optional libraries?

PR Type:

  • New Feature
  • Bugfix
  • Documentation

If you haven't finished some of the above items you can still open "Draft" PR.

Who can review?

Anyone in the community is free to review the PR once the checks have passed.
Contributor guidelines contains specific people who can review PRs to various areas.

Additional Information

  • Related to # (issue)

MaximumEntropy and others added 21 commits August 2, 2022 16:18
Signed-off-by: MaximumEntropy <[email protected]>
Signed-off-by: MaximumEntropy <[email protected]>
Signed-off-by: ericharper <[email protected]>
* [TTS] fixed wrong pronunciations.

Signed-off-by: Xuesong Yang <[email protected]>

* incremented the version number to 22.08 as @blisc suggested.

Signed-off-by: Xuesong Yang <[email protected]>

* correct cmudict versions in world-wide places.

Signed-off-by: Xuesong Yang <[email protected]>
* [TTS] incremented the version number to 22.08 in tutorials.

Signed-off-by: Xuesong Yang <[email protected]>
* Fix for RPE

Signed-off-by: MaximumEntropy <[email protected]>

* Style

Signed-off-by: MaximumEntropy <[email protected]>
Signed-off-by: MaximumEntropy <[email protected]>
Signed-off-by: MaximumEntropy <[email protected]>
Signed-off-by: MaximumEntropy <[email protected]>
Signed-off-by: MaximumEntropy <[email protected]>
Signed-off-by: MaximumEntropy <[email protected]>
Signed-off-by: MaximumEntropy <[email protected]>
Signed-off-by: MaximumEntropy <[email protected]>
@khcs khcs self-requested a review August 15, 2022 22:49
Copy link
Collaborator

@khcs khcs left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Outside of those where I left comments, we made a few changes to the files to make it work.

@khcs
Copy link
Collaborator

khcs commented Aug 15, 2022

Currently working with PP>2.
Continuing testing with different configs like bf16, O2, PP>4, different encoder/decoder RPE settings, etc.

@lgtm-com
Copy link

lgtm-com bot commented Aug 16, 2022

This pull request introduces 2 alerts when merging f5a6d03 into 06ad865 - view on LGTM.com

new alerts:

  • 2 for Unused import

Signed-off-by: MaximumEntropy <[email protected]>
@lgtm-com
Copy link

lgtm-com bot commented Aug 17, 2022

This pull request introduces 2 alerts when merging 7711620 into 4bf54b7 - view on LGTM.com

new alerts:

  • 2 for Unused import

@lgtm-com
Copy link

lgtm-com bot commented Aug 18, 2022

This pull request introduces 2 alerts when merging 187d1fd into 8845add - view on LGTM.com

new alerts:

  • 2 for Unused import

Signed-off-by: MaximumEntropy <[email protected]>
@lgtm-com
Copy link

lgtm-com bot commented Aug 18, 2022

This pull request introduces 2 alerts when merging 4526417 into 8845add - view on LGTM.com

new alerts:

  • 2 for Unused import

Signed-off-by: MaximumEntropy <[email protected]>
@lgtm-com
Copy link

lgtm-com bot commented Aug 18, 2022

This pull request introduces 2 alerts when merging 9b7a9ee into 6abfbbf - view on LGTM.com

new alerts:

  • 2 for Unused import

Signed-off-by: MaximumEntropy <[email protected]>
Signed-off-by: MaximumEntropy <[email protected]>
@lgtm-com
Copy link

lgtm-com bot commented Aug 22, 2022

This pull request introduces 2 alerts when merging bba0335 into 4a1fcc6 - view on LGTM.com

new alerts:

  • 2 for Unused import

@khcs khcs self-requested a review August 26, 2022 17:58
Copy link
Collaborator

@khcs khcs left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

ready to be merged.

@khcs khcs merged commit 4d3e335 into main Aug 26, 2022
yaoyu-33 pushed a commit that referenced this pull request Aug 26, 2022
* Enc-Dec RPE PP > 2

Signed-off-by: MaximumEntropy <[email protected]>

* Syle

Signed-off-by: MaximumEntropy <[email protected]>

* update branch

Signed-off-by: ericharper <[email protected]>

* update package info and dockerfile

Signed-off-by: ericharper <[email protected]>

* fix fastpitch export (#4676)

Signed-off-by: Jason <[email protected]>

* [TTS] fixed wrong pronunciations for r1.11. (#4677)

* [TTS] fixed wrong pronunciations.

Signed-off-by: Xuesong Yang <[email protected]>

* incremented the version number to 22.08 as @blisc suggested.

Signed-off-by: Xuesong Yang <[email protected]>

* correct cmudict versions in world-wide places.

Signed-off-by: Xuesong Yang <[email protected]>

* Fix for incorrect batch size issue while decoding (#4675)

Co-authored-by: Micha Livne <[email protected]>
Co-authored-by: Eric Harper <[email protected]>

* [TTS] incremented the version number to 22.08 in tutorials. (#4684)

* [TTS] incremented the version number to 22.08 in tutorials.

Signed-off-by: Xuesong Yang <[email protected]>

* Megatron encode function with RPE fix (#4692)

* Fix for RPE

Signed-off-by: MaximumEntropy <[email protected]>

* Style

Signed-off-by: MaximumEntropy <[email protected]>

* fix to fetch config file (#4699)

Signed-off-by: nithinraok <[email protected]>

* Fix

Signed-off-by: MaximumEntropy <[email protected]>

* Style

Signed-off-by: MaximumEntropy <[email protected]>

* Fix

Signed-off-by: MaximumEntropy <[email protected]>

* Fix rpe check

Signed-off-by: MaximumEntropy <[email protected]>

* Revert to main

Signed-off-by: MaximumEntropy <[email protected]>

* Fix

Signed-off-by: MaximumEntropy <[email protected]>

* Revert more files

Signed-off-by: MaximumEntropy <[email protected]>

* Additional changes to make PP>2 with RPE work.

* Fixes for O2

Signed-off-by: MaximumEntropy <[email protected]>

* Fixes

Signed-off-by: MaximumEntropy <[email protected]>

* Update

Signed-off-by: MaximumEntropy <[email protected]>

* Fix cross attention sync check

Signed-off-by: MaximumEntropy <[email protected]>

* Fix

Signed-off-by: MaximumEntropy <[email protected]>

* Fix

Signed-off-by: MaximumEntropy <[email protected]>

* Fix zero init

Signed-off-by: MaximumEntropy <[email protected]>

* Modify .data instead of the weight

Signed-off-by: MaximumEntropy <[email protected]>

* Add worker changes

Signed-off-by: MaximumEntropy <[email protected]>

* Set workers to 0 for validation and test

Signed-off-by: MaximumEntropy <[email protected]>

* Remove unused imports and allreduce func

Signed-off-by: MaximumEntropy <[email protected]>

* Style

Signed-off-by: MaximumEntropy <[email protected]>

* Revert workers change

Signed-off-by: MaximumEntropy <[email protected]>

Signed-off-by: MaximumEntropy <[email protected]>
Signed-off-by: ericharper <[email protected]>
Signed-off-by: Jason <[email protected]>
Signed-off-by: Xuesong Yang <[email protected]>
Signed-off-by: nithinraok <[email protected]>
Co-authored-by: ericharper <[email protected]>
Co-authored-by: Jason <[email protected]>
Co-authored-by: Xuesong Yang <[email protected]>
Co-authored-by: Rajesh Ilango <[email protected]>
Co-authored-by: Micha Livne <[email protected]>
Co-authored-by: Nithin Rao <[email protected]>
Co-authored-by: Hoo Chang Shin <[email protected]>
Co-authored-by: khcs <[email protected]>
@ericharper ericharper deleted the megatron_rpe_pp branch September 20, 2022 23:55
jubick1337 pushed a commit to jubick1337/NeMo that referenced this pull request Oct 3, 2022
* Enc-Dec RPE PP > 2

Signed-off-by: MaximumEntropy <[email protected]>

* Syle

Signed-off-by: MaximumEntropy <[email protected]>

* update branch

Signed-off-by: ericharper <[email protected]>

* update package info and dockerfile

Signed-off-by: ericharper <[email protected]>

* fix fastpitch export (NVIDIA#4676)

Signed-off-by: Jason <[email protected]>

* [TTS] fixed wrong pronunciations for r1.11. (NVIDIA#4677)

* [TTS] fixed wrong pronunciations.

Signed-off-by: Xuesong Yang <[email protected]>

* incremented the version number to 22.08 as @blisc suggested.

Signed-off-by: Xuesong Yang <[email protected]>

* correct cmudict versions in world-wide places.

Signed-off-by: Xuesong Yang <[email protected]>

* Fix for incorrect batch size issue while decoding (NVIDIA#4675)

Co-authored-by: Micha Livne <[email protected]>
Co-authored-by: Eric Harper <[email protected]>

* [TTS] incremented the version number to 22.08 in tutorials. (NVIDIA#4684)

* [TTS] incremented the version number to 22.08 in tutorials.

Signed-off-by: Xuesong Yang <[email protected]>

* Megatron encode function with RPE fix (NVIDIA#4692)

* Fix for RPE

Signed-off-by: MaximumEntropy <[email protected]>

* Style

Signed-off-by: MaximumEntropy <[email protected]>

* fix to fetch config file (NVIDIA#4699)

Signed-off-by: nithinraok <[email protected]>

* Fix

Signed-off-by: MaximumEntropy <[email protected]>

* Style

Signed-off-by: MaximumEntropy <[email protected]>

* Fix

Signed-off-by: MaximumEntropy <[email protected]>

* Fix rpe check

Signed-off-by: MaximumEntropy <[email protected]>

* Revert to main

Signed-off-by: MaximumEntropy <[email protected]>

* Fix

Signed-off-by: MaximumEntropy <[email protected]>

* Revert more files

Signed-off-by: MaximumEntropy <[email protected]>

* Additional changes to make PP>2 with RPE work.

* Fixes for O2

Signed-off-by: MaximumEntropy <[email protected]>

* Fixes

Signed-off-by: MaximumEntropy <[email protected]>

* Update

Signed-off-by: MaximumEntropy <[email protected]>

* Fix cross attention sync check

Signed-off-by: MaximumEntropy <[email protected]>

* Fix

Signed-off-by: MaximumEntropy <[email protected]>

* Fix

Signed-off-by: MaximumEntropy <[email protected]>

* Fix zero init

Signed-off-by: MaximumEntropy <[email protected]>

* Modify .data instead of the weight

Signed-off-by: MaximumEntropy <[email protected]>

* Add worker changes

Signed-off-by: MaximumEntropy <[email protected]>

* Set workers to 0 for validation and test

Signed-off-by: MaximumEntropy <[email protected]>

* Remove unused imports and allreduce func

Signed-off-by: MaximumEntropy <[email protected]>

* Style

Signed-off-by: MaximumEntropy <[email protected]>

* Revert workers change

Signed-off-by: MaximumEntropy <[email protected]>

Signed-off-by: MaximumEntropy <[email protected]>
Signed-off-by: ericharper <[email protected]>
Signed-off-by: Jason <[email protected]>
Signed-off-by: Xuesong Yang <[email protected]>
Signed-off-by: nithinraok <[email protected]>
Co-authored-by: ericharper <[email protected]>
Co-authored-by: Jason <[email protected]>
Co-authored-by: Xuesong Yang <[email protected]>
Co-authored-by: Rajesh Ilango <[email protected]>
Co-authored-by: Micha Livne <[email protected]>
Co-authored-by: Nithin Rao <[email protected]>
Co-authored-by: Hoo Chang Shin <[email protected]>
Co-authored-by: khcs <[email protected]>
Signed-off-by: Matvei Novikov <[email protected]>
hainan-xv pushed a commit to hainan-xv/NeMo that referenced this pull request Nov 29, 2022
* Enc-Dec RPE PP > 2

Signed-off-by: MaximumEntropy <[email protected]>

* Syle

Signed-off-by: MaximumEntropy <[email protected]>

* update branch

Signed-off-by: ericharper <[email protected]>

* update package info and dockerfile

Signed-off-by: ericharper <[email protected]>

* fix fastpitch export (NVIDIA#4676)

Signed-off-by: Jason <[email protected]>

* [TTS] fixed wrong pronunciations for r1.11. (NVIDIA#4677)

* [TTS] fixed wrong pronunciations.

Signed-off-by: Xuesong Yang <[email protected]>

* incremented the version number to 22.08 as @blisc suggested.

Signed-off-by: Xuesong Yang <[email protected]>

* correct cmudict versions in world-wide places.

Signed-off-by: Xuesong Yang <[email protected]>

* Fix for incorrect batch size issue while decoding (NVIDIA#4675)

Co-authored-by: Micha Livne <[email protected]>
Co-authored-by: Eric Harper <[email protected]>

* [TTS] incremented the version number to 22.08 in tutorials. (NVIDIA#4684)

* [TTS] incremented the version number to 22.08 in tutorials.

Signed-off-by: Xuesong Yang <[email protected]>

* Megatron encode function with RPE fix (NVIDIA#4692)

* Fix for RPE

Signed-off-by: MaximumEntropy <[email protected]>

* Style

Signed-off-by: MaximumEntropy <[email protected]>

* fix to fetch config file (NVIDIA#4699)

Signed-off-by: nithinraok <[email protected]>

* Fix

Signed-off-by: MaximumEntropy <[email protected]>

* Style

Signed-off-by: MaximumEntropy <[email protected]>

* Fix

Signed-off-by: MaximumEntropy <[email protected]>

* Fix rpe check

Signed-off-by: MaximumEntropy <[email protected]>

* Revert to main

Signed-off-by: MaximumEntropy <[email protected]>

* Fix

Signed-off-by: MaximumEntropy <[email protected]>

* Revert more files

Signed-off-by: MaximumEntropy <[email protected]>

* Additional changes to make PP>2 with RPE work.

* Fixes for O2

Signed-off-by: MaximumEntropy <[email protected]>

* Fixes

Signed-off-by: MaximumEntropy <[email protected]>

* Update

Signed-off-by: MaximumEntropy <[email protected]>

* Fix cross attention sync check

Signed-off-by: MaximumEntropy <[email protected]>

* Fix

Signed-off-by: MaximumEntropy <[email protected]>

* Fix

Signed-off-by: MaximumEntropy <[email protected]>

* Fix zero init

Signed-off-by: MaximumEntropy <[email protected]>

* Modify .data instead of the weight

Signed-off-by: MaximumEntropy <[email protected]>

* Add worker changes

Signed-off-by: MaximumEntropy <[email protected]>

* Set workers to 0 for validation and test

Signed-off-by: MaximumEntropy <[email protected]>

* Remove unused imports and allreduce func

Signed-off-by: MaximumEntropy <[email protected]>

* Style

Signed-off-by: MaximumEntropy <[email protected]>

* Revert workers change

Signed-off-by: MaximumEntropy <[email protected]>

Signed-off-by: MaximumEntropy <[email protected]>
Signed-off-by: ericharper <[email protected]>
Signed-off-by: Jason <[email protected]>
Signed-off-by: Xuesong Yang <[email protected]>
Signed-off-by: nithinraok <[email protected]>
Co-authored-by: ericharper <[email protected]>
Co-authored-by: Jason <[email protected]>
Co-authored-by: Xuesong Yang <[email protected]>
Co-authored-by: Rajesh Ilango <[email protected]>
Co-authored-by: Micha Livne <[email protected]>
Co-authored-by: Nithin Rao <[email protected]>
Co-authored-by: Hoo Chang Shin <[email protected]>
Co-authored-by: khcs <[email protected]>
Signed-off-by: Hainan Xu <[email protected]>
hainan-xv pushed a commit to hainan-xv/NeMo that referenced this pull request Nov 29, 2022
* Enc-Dec RPE PP > 2

Signed-off-by: MaximumEntropy <[email protected]>

* Syle

Signed-off-by: MaximumEntropy <[email protected]>

* update branch

Signed-off-by: ericharper <[email protected]>

* update package info and dockerfile

Signed-off-by: ericharper <[email protected]>

* fix fastpitch export (NVIDIA#4676)

Signed-off-by: Jason <[email protected]>

* [TTS] fixed wrong pronunciations for r1.11. (NVIDIA#4677)

* [TTS] fixed wrong pronunciations.

Signed-off-by: Xuesong Yang <[email protected]>

* incremented the version number to 22.08 as @blisc suggested.

Signed-off-by: Xuesong Yang <[email protected]>

* correct cmudict versions in world-wide places.

Signed-off-by: Xuesong Yang <[email protected]>

* Fix for incorrect batch size issue while decoding (NVIDIA#4675)

Co-authored-by: Micha Livne <[email protected]>
Co-authored-by: Eric Harper <[email protected]>

* [TTS] incremented the version number to 22.08 in tutorials. (NVIDIA#4684)

* [TTS] incremented the version number to 22.08 in tutorials.

Signed-off-by: Xuesong Yang <[email protected]>

* Megatron encode function with RPE fix (NVIDIA#4692)

* Fix for RPE

Signed-off-by: MaximumEntropy <[email protected]>

* Style

Signed-off-by: MaximumEntropy <[email protected]>

* fix to fetch config file (NVIDIA#4699)

Signed-off-by: nithinraok <[email protected]>

* Fix

Signed-off-by: MaximumEntropy <[email protected]>

* Style

Signed-off-by: MaximumEntropy <[email protected]>

* Fix

Signed-off-by: MaximumEntropy <[email protected]>

* Fix rpe check

Signed-off-by: MaximumEntropy <[email protected]>

* Revert to main

Signed-off-by: MaximumEntropy <[email protected]>

* Fix

Signed-off-by: MaximumEntropy <[email protected]>

* Revert more files

Signed-off-by: MaximumEntropy <[email protected]>

* Additional changes to make PP>2 with RPE work.

* Fixes for O2

Signed-off-by: MaximumEntropy <[email protected]>

* Fixes

Signed-off-by: MaximumEntropy <[email protected]>

* Update

Signed-off-by: MaximumEntropy <[email protected]>

* Fix cross attention sync check

Signed-off-by: MaximumEntropy <[email protected]>

* Fix

Signed-off-by: MaximumEntropy <[email protected]>

* Fix

Signed-off-by: MaximumEntropy <[email protected]>

* Fix zero init

Signed-off-by: MaximumEntropy <[email protected]>

* Modify .data instead of the weight

Signed-off-by: MaximumEntropy <[email protected]>

* Add worker changes

Signed-off-by: MaximumEntropy <[email protected]>

* Set workers to 0 for validation and test

Signed-off-by: MaximumEntropy <[email protected]>

* Remove unused imports and allreduce func

Signed-off-by: MaximumEntropy <[email protected]>

* Style

Signed-off-by: MaximumEntropy <[email protected]>

* Revert workers change

Signed-off-by: MaximumEntropy <[email protected]>

Signed-off-by: MaximumEntropy <[email protected]>
Signed-off-by: ericharper <[email protected]>
Signed-off-by: Jason <[email protected]>
Signed-off-by: Xuesong Yang <[email protected]>
Signed-off-by: nithinraok <[email protected]>
Co-authored-by: ericharper <[email protected]>
Co-authored-by: Jason <[email protected]>
Co-authored-by: Xuesong Yang <[email protected]>
Co-authored-by: Rajesh Ilango <[email protected]>
Co-authored-by: Micha Livne <[email protected]>
Co-authored-by: Nithin Rao <[email protected]>
Co-authored-by: Hoo Chang Shin <[email protected]>
Co-authored-by: khcs <[email protected]>
Signed-off-by: Hainan Xu <[email protected]>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

7 participants