Speeds up copying of necessary artifact files with SaveRestoreConnector #9682
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
Previously, the SaveRestoreConnector would copy and untar entire checkpoints just to copy out a tokenizer. For models in the >100GB, this led to timeouts since only rank=0 did this work, while other ranks moved on and waited at an all-gather barrier (observed NCCL timeout at 10min). Signed-off-by: Terry Kong <terryk@nvidia.com>
Signed-off-by: Terry Kong <terryk@nvidia.com>
Signed-off-by: terrykong <terrykong@users.noreply.github.com> Signed-off-by: Terry Kong <terryk@nvidia.com>
Signed-off-by: Terry Kong <terryk@nvidia.com>
Signed-off-by: Terry Kong <terryk@nvidia.com>
Signed-off-by: Terry Kong <terryk@nvidia.com>
Signed-off-by: Terry Kong <terryk@nvidia.com>
Signed-off-by: terrykong <terrykong@users.noreply.github.com>
Signed-off-by: Terry Kong <terryk@nvidia.com>
titu1994
previously approved these changes
Jul 11, 2024
|
I'm looking into the test failures. Very bizarre since running the complete suite locally hangs and I cannot reproduce the failures by running the failed cases individually |
Signed-off-by: Terry Kong <terryk@nvidia.com>
Signed-off-by: terrykong <terrykong@users.noreply.github.com>
|
I figured out the test failure issue. It turned out in my unit test I |
ertkonuk
pushed a commit
that referenced
this pull request
Jul 19, 2024
…or (#9682) * Speeds up copying of neccesary artifact files with SaveRestoreConnector Previously, the SaveRestoreConnector would copy and untar entire checkpoints just to copy out a tokenizer. For models in the >100GB, this led to timeouts since only rank=0 did this work, while other ranks moved on and waited at an all-gather barrier (observed NCCL timeout at 10min). Signed-off-by: Terry Kong <terryk@nvidia.com> * cleanup Signed-off-by: Terry Kong <terryk@nvidia.com> * black formatting Signed-off-by: Terry Kong <terryk@nvidia.com> * Apply isort and black reformatting Signed-off-by: terrykong <terrykong@users.noreply.github.com> Signed-off-by: Terry Kong <terryk@nvidia.com> * restoring logic to previous tempdir logic Signed-off-by: Terry Kong <terryk@nvidia.com> * nlp overrides too Signed-off-by: Terry Kong <terryk@nvidia.com> * respect return_config Signed-off-by: Terry Kong <terryk@nvidia.com> * some unit tests Signed-off-by: Terry Kong <terryk@nvidia.com> * nodbg Signed-off-by: Terry Kong <terryk@nvidia.com> * Apply isort and black reformatting Signed-off-by: terrykong <terrykong@users.noreply.github.com> * correct typing Signed-off-by: Terry Kong <terryk@nvidia.com> * Fixes directory issue Signed-off-by: Terry Kong <terryk@nvidia.com> * Apply isort and black reformatting Signed-off-by: terrykong <terrykong@users.noreply.github.com> --------- Signed-off-by: Terry Kong <terryk@nvidia.com> Signed-off-by: terrykong <terrykong@users.noreply.github.com> Co-authored-by: terrykong <terrykong@users.noreply.github.com> Co-authored-by: Eric Harper <complex451@gmail.com> Signed-off-by: Tugrul Konuk <ertkonuk@gmail.com>
8 tasks
tonyjie
pushed a commit
to tonyjie/NeMo
that referenced
this pull request
Jul 24, 2024
…or (NVIDIA#9682) * Speeds up copying of neccesary artifact files with SaveRestoreConnector Previously, the SaveRestoreConnector would copy and untar entire checkpoints just to copy out a tokenizer. For models in the >100GB, this led to timeouts since only rank=0 did this work, while other ranks moved on and waited at an all-gather barrier (observed NCCL timeout at 10min). Signed-off-by: Terry Kong <terryk@nvidia.com> * cleanup Signed-off-by: Terry Kong <terryk@nvidia.com> * black formatting Signed-off-by: Terry Kong <terryk@nvidia.com> * Apply isort and black reformatting Signed-off-by: terrykong <terrykong@users.noreply.github.com> Signed-off-by: Terry Kong <terryk@nvidia.com> * restoring logic to previous tempdir logic Signed-off-by: Terry Kong <terryk@nvidia.com> * nlp overrides too Signed-off-by: Terry Kong <terryk@nvidia.com> * respect return_config Signed-off-by: Terry Kong <terryk@nvidia.com> * some unit tests Signed-off-by: Terry Kong <terryk@nvidia.com> * nodbg Signed-off-by: Terry Kong <terryk@nvidia.com> * Apply isort and black reformatting Signed-off-by: terrykong <terrykong@users.noreply.github.com> * correct typing Signed-off-by: Terry Kong <terryk@nvidia.com> * Fixes directory issue Signed-off-by: Terry Kong <terryk@nvidia.com> * Apply isort and black reformatting Signed-off-by: terrykong <terrykong@users.noreply.github.com> --------- Signed-off-by: Terry Kong <terryk@nvidia.com> Signed-off-by: terrykong <terrykong@users.noreply.github.com> Co-authored-by: terrykong <terrykong@users.noreply.github.com> Co-authored-by: Eric Harper <complex451@gmail.com>
akoumpa
pushed a commit
that referenced
this pull request
Jul 25, 2024
…or (#9682) * Speeds up copying of neccesary artifact files with SaveRestoreConnector Previously, the SaveRestoreConnector would copy and untar entire checkpoints just to copy out a tokenizer. For models in the >100GB, this led to timeouts since only rank=0 did this work, while other ranks moved on and waited at an all-gather barrier (observed NCCL timeout at 10min). Signed-off-by: Terry Kong <terryk@nvidia.com> * cleanup Signed-off-by: Terry Kong <terryk@nvidia.com> * black formatting Signed-off-by: Terry Kong <terryk@nvidia.com> * Apply isort and black reformatting Signed-off-by: terrykong <terrykong@users.noreply.github.com> Signed-off-by: Terry Kong <terryk@nvidia.com> * restoring logic to previous tempdir logic Signed-off-by: Terry Kong <terryk@nvidia.com> * nlp overrides too Signed-off-by: Terry Kong <terryk@nvidia.com> * respect return_config Signed-off-by: Terry Kong <terryk@nvidia.com> * some unit tests Signed-off-by: Terry Kong <terryk@nvidia.com> * nodbg Signed-off-by: Terry Kong <terryk@nvidia.com> * Apply isort and black reformatting Signed-off-by: terrykong <terrykong@users.noreply.github.com> * correct typing Signed-off-by: Terry Kong <terryk@nvidia.com> * Fixes directory issue Signed-off-by: Terry Kong <terryk@nvidia.com> * Apply isort and black reformatting Signed-off-by: terrykong <terrykong@users.noreply.github.com> --------- Signed-off-by: Terry Kong <terryk@nvidia.com> Signed-off-by: terrykong <terrykong@users.noreply.github.com> Co-authored-by: terrykong <terrykong@users.noreply.github.com> Co-authored-by: Eric Harper <complex451@gmail.com> Signed-off-by: Alexandros Koumparoulis <akoumparouli@nvidia.com>
malay-nagda
pushed a commit
to malay-nagda/NeMo
that referenced
this pull request
Jul 26, 2024
…or (NVIDIA#9682) * Speeds up copying of neccesary artifact files with SaveRestoreConnector Previously, the SaveRestoreConnector would copy and untar entire checkpoints just to copy out a tokenizer. For models in the >100GB, this led to timeouts since only rank=0 did this work, while other ranks moved on and waited at an all-gather barrier (observed NCCL timeout at 10min). Signed-off-by: Terry Kong <terryk@nvidia.com> * cleanup Signed-off-by: Terry Kong <terryk@nvidia.com> * black formatting Signed-off-by: Terry Kong <terryk@nvidia.com> * Apply isort and black reformatting Signed-off-by: terrykong <terrykong@users.noreply.github.com> Signed-off-by: Terry Kong <terryk@nvidia.com> * restoring logic to previous tempdir logic Signed-off-by: Terry Kong <terryk@nvidia.com> * nlp overrides too Signed-off-by: Terry Kong <terryk@nvidia.com> * respect return_config Signed-off-by: Terry Kong <terryk@nvidia.com> * some unit tests Signed-off-by: Terry Kong <terryk@nvidia.com> * nodbg Signed-off-by: Terry Kong <terryk@nvidia.com> * Apply isort and black reformatting Signed-off-by: terrykong <terrykong@users.noreply.github.com> * correct typing Signed-off-by: Terry Kong <terryk@nvidia.com> * Fixes directory issue Signed-off-by: Terry Kong <terryk@nvidia.com> * Apply isort and black reformatting Signed-off-by: terrykong <terrykong@users.noreply.github.com> --------- Signed-off-by: Terry Kong <terryk@nvidia.com> Signed-off-by: terrykong <terrykong@users.noreply.github.com> Co-authored-by: terrykong <terrykong@users.noreply.github.com> Co-authored-by: Eric Harper <complex451@gmail.com> Signed-off-by: Malay Nagda <malayn@malayn-mlt.client.nvidia.com>
monica-sekoyan
pushed a commit
that referenced
this pull request
Oct 14, 2024
…or (#9682) * Speeds up copying of neccesary artifact files with SaveRestoreConnector Previously, the SaveRestoreConnector would copy and untar entire checkpoints just to copy out a tokenizer. For models in the >100GB, this led to timeouts since only rank=0 did this work, while other ranks moved on and waited at an all-gather barrier (observed NCCL timeout at 10min). Signed-off-by: Terry Kong <terryk@nvidia.com> * cleanup Signed-off-by: Terry Kong <terryk@nvidia.com> * black formatting Signed-off-by: Terry Kong <terryk@nvidia.com> * Apply isort and black reformatting Signed-off-by: terrykong <terrykong@users.noreply.github.com> Signed-off-by: Terry Kong <terryk@nvidia.com> * restoring logic to previous tempdir logic Signed-off-by: Terry Kong <terryk@nvidia.com> * nlp overrides too Signed-off-by: Terry Kong <terryk@nvidia.com> * respect return_config Signed-off-by: Terry Kong <terryk@nvidia.com> * some unit tests Signed-off-by: Terry Kong <terryk@nvidia.com> * nodbg Signed-off-by: Terry Kong <terryk@nvidia.com> * Apply isort and black reformatting Signed-off-by: terrykong <terrykong@users.noreply.github.com> * correct typing Signed-off-by: Terry Kong <terryk@nvidia.com> * Fixes directory issue Signed-off-by: Terry Kong <terryk@nvidia.com> * Apply isort and black reformatting Signed-off-by: terrykong <terrykong@users.noreply.github.com> --------- Signed-off-by: Terry Kong <terryk@nvidia.com> Signed-off-by: terrykong <terrykong@users.noreply.github.com> Co-authored-by: terrykong <terrykong@users.noreply.github.com> Co-authored-by: Eric Harper <complex451@gmail.com>
hainan-xv
pushed a commit
to hainan-xv/NeMo
that referenced
this pull request
Nov 5, 2024
…or (NVIDIA#9682) * Speeds up copying of neccesary artifact files with SaveRestoreConnector Previously, the SaveRestoreConnector would copy and untar entire checkpoints just to copy out a tokenizer. For models in the >100GB, this led to timeouts since only rank=0 did this work, while other ranks moved on and waited at an all-gather barrier (observed NCCL timeout at 10min). Signed-off-by: Terry Kong <terryk@nvidia.com> * cleanup Signed-off-by: Terry Kong <terryk@nvidia.com> * black formatting Signed-off-by: Terry Kong <terryk@nvidia.com> * Apply isort and black reformatting Signed-off-by: terrykong <terrykong@users.noreply.github.com> Signed-off-by: Terry Kong <terryk@nvidia.com> * restoring logic to previous tempdir logic Signed-off-by: Terry Kong <terryk@nvidia.com> * nlp overrides too Signed-off-by: Terry Kong <terryk@nvidia.com> * respect return_config Signed-off-by: Terry Kong <terryk@nvidia.com> * some unit tests Signed-off-by: Terry Kong <terryk@nvidia.com> * nodbg Signed-off-by: Terry Kong <terryk@nvidia.com> * Apply isort and black reformatting Signed-off-by: terrykong <terrykong@users.noreply.github.com> * correct typing Signed-off-by: Terry Kong <terryk@nvidia.com> * Fixes directory issue Signed-off-by: Terry Kong <terryk@nvidia.com> * Apply isort and black reformatting Signed-off-by: terrykong <terrykong@users.noreply.github.com> --------- Signed-off-by: Terry Kong <terryk@nvidia.com> Signed-off-by: terrykong <terrykong@users.noreply.github.com> Co-authored-by: terrykong <terrykong@users.noreply.github.com> Co-authored-by: Eric Harper <complex451@gmail.com> Signed-off-by: Hainan Xu <hainanx@nvidia.com>
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.
What does this PR do ?
Previously, the SaveRestoreConnector would copy and untar entire
checkpoints just to copy out a tokenizer. For models in the >100GB, this
led to timeouts since only rank=0 did this work, while other ranks moved
on and waited at an all-gather barrier (observed NCCL timeout at 10min).
Revision of #9299 that keeps most of the core logic surrounding tempdir creation