[autoparallel] Hook all meta information on ResNet nodes for auto activation checkpoint#2248
Merged
super-dainiu merged 57 commits intohpcaitech:debug/ckpt-autoparallelfrom Jan 2, 2023
Conversation
Merge ColossalAI
Daily merge
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
2 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What's New?
In this PR, I manage to hook all meta information on ResNet nodes so that the graph module could be prepared for the auto-activation checkpoint solver. I introduce a new pass
comm_metainfo_passto process the meta information of communication nodes such asruntime_applyandruntime_comm_spec_apply. I also polish some code in the meta-information generator and fix some small bugs along the way.NOTE:
In this PR, I don't modify some node handlers to turn them into
MetaInfoNodeHandlerorMetaInfoModuleHandleras some of the operations haven't been patched, I will work on it in the next year (today is the last day of the year 2022 lol). I did the test locally and verified that the meta-information of ResNet could be successfully hooked on nodes.Last PR for 2022, wish you guys all the best and have a lovely year in 2023.