contramundum53

Any updates on this issue?

Some benchmark results: sigopt benchmarks (deterministic function):     HPOBench (stochastic function, discrete space; Note that the objective function is very ill-scaled, so taking log will help for...

@nabenabe0928 @y0z Could you review this PR?

Note that our sampler is designed for real-world problems, and real-world hyperparameter optimization problems have different characteristics from benchmark functions like `sigopt` functions. Most notably, real-world problems tend to be...

Also we do not publicly expose the interface for getting posterior distribution of the surrogate model, but you can mimic the logic in `GPSampler` (fitting the model is deterministic) and...

@nabenabe0928 I applied all comments. PTAL.

@nabenabe0928 I applied your comments, PTAL!

@andreyvelich @tenzen-y I'm glad that you are considering integration of Optuna's `GPSampler` into katib! `GPSampler` is planned to be released in Optuna v3.6, our next minor release. It is still...

Sorry no, that will depend on many factors, such as how much the new feature is used, how many feature requests come, how many bugs we encounter, and so on.

Why did we fix PyTorch version to 1.11? If we could remove that constraint, maybe we could resolve this problem?