Earlier versions you could do logging like this:
python train.py --num-workers 2 --env-id PongDeterministic-v3 --log-dir /tmp/pong
then starting Tensorboard and monitor progress.
It's probably just me, but with baselines.run I can't find a kwarg to specify the log directory/name or activate logging. And I haven't found any recent posts on Tensorboard / openai either.
There is a logger.py file, and a results_plotter.py. Probably I activate the logger with some kwarg, but how?
And is Tensorboard totally out of the picture now?
Kind regards
you need to set the OPENAI_LOG_FORMAT and OPENAI_LOGDIR environment variables, e.g.,
export OPENAI_LOG_FORMAT='stdout,log,csv,tensorboard'
export OPENAI_LOGDIR=....your_path...
exactly as @jramak has said :) I'll keep this open to add instructions to README
Thanks. That would be very helpful!
I really appreciate the approachability and helpfulness of you guys!
On 4 September 2018 at 22:33:32, pzhokhov ([email protected]) wrote:
exactly as @jramak has said :) I'll keep this open to add instructions on that into README
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub, or mute the thread.
@pzhokhov I created a PR to include this in the README:
how do I link a PR to this thread rather than just refer to it in the comment?
nevermind, got it :)
Most helpful comment
you need to set the OPENAI_LOG_FORMAT and OPENAI_LOGDIR environment variables, e.g.,
export OPENAI_LOG_FORMAT='stdout,log,csv,tensorboard'export OPENAI_LOGDIR=....your_path...