Skip to content

Conversation

@chi-yelp
Copy link
Contributor

@chi-yelp chi-yelp commented May 19, 2025

Changes

  • Keep minExecutors as Spark default 0, instead of using spark.yelp.dra.minExecutorRatio to calculate
    • This avoids additional costs from idle executors, especially for gpu training
  • Set spark.executor.instances as initialExecutors instead of minExecutors
    • The default value of initialExecutors is minExecutors

Test

Tested with paasta spark-run and spark-tools in jupyter by installed the whl.

@chi-yelp chi-yelp force-pushed the u/chi/revise_dra_default_settings branch from adfb92b to 1ce188d Compare May 19, 2025 18:07
@chi-yelp chi-yelp marked this pull request as ready for review May 19, 2025 18:12
@chi-yelp chi-yelp requested review from austinzh and timma-yelp May 19, 2025 18:12
@chi-yelp chi-yelp force-pushed the u/chi/revise_dra_default_settings branch from 1ce188d to bba400f Compare May 19, 2025 19:10
@chi-yelp chi-yelp merged commit 05dcb28 into master May 20, 2025
1 check passed
@chi-yelp chi-yelp deleted the u/chi/revise_dra_default_settings branch May 20, 2025 10:24
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants