Florin Andrei
Florin Andrei
@AlexM4 That is a MASSIVE improvement!
@alpotapov Have you found a better / faster version of do_nms() ?
Right after I wrote the bug report, it started to print messages in the notebook: ``` Features: 3000/524287IOPub message rate exceeded. The notebook server will temporarily stop sending output to...
I'm dealing with a time crunch right now, I'm not sure if I can help. I'll add a comment here if that changes. Sorry!
This would be a really nice feature to have.
It could also be the other way 'round: since most people have no simple way to find and track these objects, because popular apps don't have them, you'd expect a...
@c-bata If I specify the TPESampler explicitly in Optuna, it gets stuck very quickly in a local extreme and it does not make much progress after that, on this problem....
@c-bata @nomuramasahir0 I looked at the CMA-ES sampler in the Optuna code and compared it with my code. I've debugged the `CmaEsSampler()` class while invoking it from Optuna, and peeked...
I've tried (1+1)-CMA-ES. It's tricky to do a direct comparison with CMA-ES with Margin, because the (1+1) code is quite different (generation count is not exposed, etc) and I didn't...
@nomuramasahir0 Thank you for your explanation. That indeed makes sense of what I'm seeing here. I will add the plain CMA code to my stash of code snippets for feature...