Deprecate ft-restart-policy (min-healthy) #259
Draft
hexinw-nvidia wants to merge 1 commit intoNVIDIA:mainfrom
Draft
Deprecate ft-restart-policy (min-healthy) #259hexinw-nvidia wants to merge 1 commit intoNVIDIA:mainfrom
hexinw-nvidia wants to merge 1 commit intoNVIDIA:mainfrom
Conversation
…example The previous ft-restart-policy design exposed two separate restart levels: InJob (launcher) and InProcess. That was confusing for users, who expect a single restart per training cycle—either some processes stay up across the restart or they are restarted for the next cycle. The current logic does not match that model and needs a proper re-design. - Deprecate --ft-restart-policy: only any-failed is supported; mark option deprecated and remove min-healthy implementation. - Remove min-healthy code path (_invoke_run_with_min_healthy_policy) and always use any-failed behavior; set upscaling_enabled=True. - Remove in_job_and_in_process example (script, Python example, and doc) and references; document that injob+inprocess integration is under re-evaluation. - Update docs (usage_guide, inprocess integration, examples toctree) and inprocess usage_guide to drop min-healthy and example references. Removing this code simplifies the codebase and makes the intended restart model easier to reason about before a future re-design.
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Remove inprocess+injob example
The previous ft-restart-policy design exposed two separate restart levels: InJob (launcher) and InProcess. That was confusing for users, who expect a single restart per training cycle—either some processes stay up across the restart or they are restarted for the next cycle. The current logic does not match that model and needs a proper re-design.
Removing this code simplifies the codebase and makes the intended restart model easier to reason about before a future re-design.