FLAML
FLAML copied to clipboard
Update Optuna
setup.py specifies that the required version of Optuna to be 2.8.0. However, according to ray
docs (check out the second warning on this page) :
Please note that with Optuna 2.10.0 and earlier default
MOTPESampler
/TPESampler
suffer from performance issues when dealing with a large number of completed trials (approx. >100). This will manifest as a delay when suggesting new configurations. This is an Optuna issue and may be fixed in a future Optuna release.
In my own experience, I have also observed the slowdown they mention.
Is there any reason for this version requirement for Optuna? And, if yes, how hard would it be update to the latest version?
Before transitioning to newer versions, it typically involves a substantial amount of effort. I'm planning to submit a pull request to update Optuna to version 2.10.0, but the rationale behind this decision needs to be verified by Microsoft employees who are actively involved with FLAML.
Hi @Programmer-RD-AI , as @bbudescu quoted, Optuna 2.10.0 also has the performance issue. I've upgraded Optuna to support latest version in PR #1321 . Could you help review? Thanks.