You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi! Good question. That isn't built in explicitly, but what I usually to do test is make copy of the script that has build_and_train in it, call it script_test.py, and inside that one I'll manually change whatever configurations I'm testing, and usually also set the config["runner"]["n_steps"] and config["runner"]["log_interval_steps"] to be small values, so that the diagnostics come back fast. Just remove the variant loading, and use something simple for the affinity code like: 0slt_1gpu_4cpu (if you have 4 cpus) or 0slot_0gpu_4cpu_4cpr if you have 4 cpus but no gpus.
Hi! Thanks for the response. I got you, but still, building an explicit build and test may lower the difficulty for new comers (like me).
Actually, I finally refered to the code from https://github.com/petrikvladimir/rlpyt_examples, where there are build and test function and simple examples on simple environments like cartpole.
Do you think the usage of rlpyt in this repo is okay?
Like this nice work of rl codebase. It's nice and pretty!
I wonder if there is any evaluation examples? I mean we have
build and train
, but do we havebuild and test
?The text was updated successfully, but these errors were encountered: