[rllib] move evaluation to trainer.step() such that the result is properly logged#12708
[rllib] move evaluation to trainer.step() such that the result is properly logged#12708sven1977 merged 2 commits intoray-project:masterfrom
Conversation
sven1977
left a comment
There was a problem hiding this comment.
Looks fine for me.
@Maltimore Could you fix the broken eval test case and run the ci/travis/format.sh (LINT) before pushing?
Thanks!
|
I ran
|
…gged even when training with trainer.train() instead of tune.run() lint bugfix: need to increment self._iteration in if condition [rllib] evaluation: simplify logic in if-condition
|
@sven1977 |
sven1977
left a comment
There was a problem hiding this comment.
Thanks for this fix @Maltimore !
Looks good to me.
…t is properly logged (ray-project#12708)" This reverts commit 59010e2.
Move evaluation to
trainer.step()such that the result is properly logged even when training withtrainer.train()instead oftune.run()Why are these changes needed?
When training with
trainer.train()instead oftune.run(), the results of the evaluation (evaluation metrics) are not written to disk (e. g. to the progress.csv).This moves the evaluation code to
trainer.step(), which ensures that the evaluation metrics are included in progress.csv and other output files.Related issue number
Checks
scripts/format.shto lint the changes in this PR. --> yes but this errors with a flake8 error unrelated to the codeI only tested this PR in my own experiments.