Open cathywu opened 6 years ago
Rllib team is working on this
I updated the above message to reflect a running version of Flow using ray==0.5.0
(without the error messages).
This looks better but I suppose it could still be less verbose? Worth figuring out on whose end the MeanOfEmptySlice thing is happening
The first run of an experiment after installation is extremely verbose, which may be a poor first experience. How much of this can we remove? How much should we keep but with an optional verbosity parameter?
Here is an example of the output from Flow in just the first iteration of training: