Improved random seed configuration for Lhotse dataloaders with docs #9001
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
Signed-off-by: Piotr Żelasko <petezor@gmail.com>
Signed-off-by: Piotr Żelasko <petezor@gmail.com>
zhehuaichen
reviewed
Apr 23, 2024
docs/source/asr/datasets.rst
Outdated
|
|
||
| * This setup guarantees 100% dataloading reproducibility. | ||
|
|
||
| * Resuming training without changing of the ``seed`` value will cause the model to train on data it has already seen. For large data setups, not managing the ``seed`` may cause the model to never be trained on a majority of data. This is why this mode is not the default. |
There was a problem hiding this comment.
what's the recommended practice on managing the "seed" on large data setup?
There was a problem hiding this comment.
Generally every time you resume, you'd provide a different value to model.train_ds.seed=<val>. A true-enough random seed can be obtained on most systems by reading /dev/urandom, e.g. uint32 seed: RSEED=$(od -An -N4 -tu4 < /dev/urandom | tr -d ' '). If you have some sort of "launcher script" that queues multiple jobs, this would be the right place to use this. Let me update the docs with this example.
Ideally we'd be able to automate this seed management thing by keeping some state in the checkpoints, but at this point it'd be a scope creep.
Signed-off-by: Piotr Żelasko <petezor@gmail.com>
erastorgueva-nv
requested changes
Apr 23, 2024
Co-authored-by: Elena Rastorgueva <80532067+erastorgueva-nv@users.noreply.github.com> Signed-off-by: Piotr Żelasko <petezor@gmail.com>
|
Thanks @erastorgueva-nv, for some reason I thought it's required by sphinx. It should look much better indeed! |
erastorgueva-nv
approved these changes
Apr 24, 2024
|
Docs LGTM |
suiyoubi
pushed a commit
that referenced
this pull request
May 2, 2024
…9001) * Improving RNG seeding with Lhotse dataloading Signed-off-by: Piotr Żelasko <petezor@gmail.com> * Fix Signed-off-by: Piotr Żelasko <petezor@gmail.com> * Add documentation about random seeds Signed-off-by: Piotr Żelasko <petezor@gmail.com> * Add doc about managing random seed Signed-off-by: Piotr Żelasko <petezor@gmail.com> * Apply suggestions from code review Co-authored-by: Elena Rastorgueva <80532067+erastorgueva-nv@users.noreply.github.com> Signed-off-by: Piotr Żelasko <petezor@gmail.com> --------- Signed-off-by: Piotr Żelasko <petezor@gmail.com> Co-authored-by: Elena Rastorgueva <80532067+erastorgueva-nv@users.noreply.github.com> Signed-off-by: Ao Tang <aot@nvidia.com>
rohitrango
pushed a commit
to rohitrango/NeMo
that referenced
this pull request
Jun 25, 2024
…VIDIA#9001) * Improving RNG seeding with Lhotse dataloading Signed-off-by: Piotr Żelasko <petezor@gmail.com> * Fix Signed-off-by: Piotr Żelasko <petezor@gmail.com> * Add documentation about random seeds Signed-off-by: Piotr Żelasko <petezor@gmail.com> * Add doc about managing random seed Signed-off-by: Piotr Żelasko <petezor@gmail.com> * Apply suggestions from code review Co-authored-by: Elena Rastorgueva <80532067+erastorgueva-nv@users.noreply.github.com> Signed-off-by: Piotr Żelasko <petezor@gmail.com> --------- Signed-off-by: Piotr Żelasko <petezor@gmail.com> Co-authored-by: Elena Rastorgueva <80532067+erastorgueva-nv@users.noreply.github.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 ?
Collection: ASR
Changelog
Usage
# Add a code snippet demonstrating how to use thisJenkins CI
The Jenkins CI system has been replaced by GitHub Actions self-hosted runners.
There's no need to comment
jenkinson the PR to trigger Jenkins CI.The GitHub Actions CI will run automatically when the PR is opened.
To run CI on an untrusted fork, a NeMo user with write access must click "Approve and run".
Before your PR is "Ready for review"
Pre checks:
PR Type:
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