An approach to sync between the nodes during the checkpoint copy activity #14
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
hosseinsarshar
changed the title
|
I closed the previous PR #13 as it was using an old approach. |
bvrockwell
reviewed
Sep 4, 2024
nemo/launch.sh
Outdated
| @@ -55,19 +58,20 @@ export ADDITIONAL_ARGS="++model.micro_batch_size=$MICRO_BATCH ++trainer.max_step | |||
| # == construct job launch command == | |||
|
|
|||
| # create base job launch command | |||
| export LAUNCH_CMD="git clone https://github.com/hosseinsarshar/dist-training-vertex.git &&" | |||
| export LAUNCH_CMD="git clone -b sync-copy-mpi https://github.com/hosseinsarshar/dist-training-vertex.git &&" | |||
There was a problem hiding this comment.
just need to remove the branch reference
There was a problem hiding this comment.
fixed in this commit e0b885d
bvrockwell
reviewed
Sep 4, 2024
There was a problem hiding this comment.
pushed these by accident i think
There was a problem hiding this comment.
yes - removed them in this commit 8920249
bvrockwell
reviewed
Sep 4, 2024
nemo/utils/model_copy.sh
Outdated
| @@ -0,0 +1,31 @@ | |||
| #!/bin/bash | |||
There was a problem hiding this comment.
I don't know why, but it might be better to name the python script and bash script differently haha - i had to double take.
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Outdated suggestions cannot be applied.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.
Things to expect in this PR:
nemo/utils/model_copy.shthat launchesmodel_copy.py- a distributed approach to sync between ranks when rank 0 is copying the checkpointdist.barrier()to sync between the nodes - I set a manual timeout to fail if it takes longer than 30 min as it might get stuck for ever.Link to a test job: nemo_llama3-70b_continual-pretraining_8