[trainer/deepspeed] load_best_model (reimplement re-init)#17151
Merged
Conversation
4 tasks
|
The documentation is not available anymore as the PR was closed or merged. |
Contributor
|
when will this pr be merged ? |
Contributor
Author
|
We needed to wait for a new deepspeed release, which I see has happened already, so yes, we can merge this shortly. |
stas00
changed the title
3 tasks
Narsil
pushed a commit
to Narsil/transformers
that referenced
this pull request
Jun 7, 2022
…e#17151) * [trainer/deepspeed] load_best_model * to sync with DS PR huggingface#1947 * simplify * rework load_best_model test * cleanup * bump deepspeed>=0.6.5 Co-authored-by: Olatunji Ruwase <olruwase@microsoft.com>
elusenji
pushed a commit
to elusenji/transformers
that referenced
this pull request
Jun 12, 2022
…e#17151) * [trainer/deepspeed] load_best_model * to sync with DS PR huggingface#1947 * simplify * rework load_best_model test * cleanup * bump deepspeed>=0.6.5 Co-authored-by: Olatunji Ruwase <olruwase@microsoft.com>
amyeroberts
pushed a commit
to amyeroberts/transformers
that referenced
this pull request
Jun 16, 2022
…e#17151) * [trainer/deepspeed] load_best_model * to sync with DS PR huggingface#1947 * simplify * rework load_best_model test * cleanup * bump deepspeed>=0.6.5 Co-authored-by: Olatunji Ruwase <olruwase@microsoft.com>
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
This PR fixes #17114
The
deepspeed_reinithack proved to not always work, as some stored args appear to be either stale or wrong (e.g. the optimizer can be a deepspeed outer optimizer which shouldn't be the case), so trying to just do a full init from scratch instead.And then there was an issue on the deepspeed side with model getting deepspeed hooks added multiple time which was breaking everything. Fixed in deepspeedai/DeepSpeed#1947.
I spent many hours trying to reproduce the problem in the usual way via examples scripts to make a test, but alas, it just won't fail in the right places. So I ended up re-implemented
test_load_best_modelusing code derived from @base-y's repro example. So super appreciating having their script.Blocking events
Fixes: #17114