[RLlib] [docs] CLI user guide #29661
Conversation
Signed-off-by: Max Pumperla <[email protected]>
maxpumperla
requested review from
sven1977,
gjoliver,
avnishn,
ArturNiederfahrenhorst,
smorad,
kouroshHakha,
krfricke and
a team
as code owners
| @@ -1,4 +1,3 @@ | |||
| # config for quick cli testing. does not give meaningful results. | |||
There was a problem hiding this comment.
Hmm, True, though, should we make this example not learn to the end (e.g. 150 reward)? It's pretty quick with SimpleQ. Otherwise, it's not really a "tuned" example :)
There was a problem hiding this comment.
yeah, good point. maybe we should move these files to another folder or so? In this case you're absolutely right, but I really want to have a couple of test files that return in a manner of seconds, not minutes - and give me a checkpoint to evaluate.
There was a problem hiding this comment.
@sven1977 ok, I reset the reward for these examples back to 150, makes sense. if we need other, quicker configs, we can add them later.
| @@ -0,0 +1,336 @@ | |||
| --- | |||
There was a problem hiding this comment.
Awesomeness! :) Dumb question. Why not rst, but md?
There was a problem hiding this comment.
ha, thanks. well, a) I personally prefer markdown, b) the margin directive doesn't work in rst, and c) the yaml header in the file makes the whole thing an executable format.
It's essentially a jupyter notebook in disguise. The serve team almost exclusively used md by now, and I just wanted to give a reference example for you guys to work off of.
There was a problem hiding this comment.
LGTM. Thanks for this great PR @maxpumperla ! The CLI is really dope now.
Just 2 questions & nits, then happy to merge.
Signed-off-by: Max Pumperla <[email protected]>
Signed-off-by: Max Pumperla <[email protected]>
Signed-off-by: Max Pumperla <[email protected]>
Signed-off-by: Weichen Xu <[email protected]>
Best to simply check the page itself: https://ray--29661.org.readthedocs.build/en/29661/rllib/rllib-cli.html