Minor changes - Githubissues

automl / mf-prior-bench

A collection of multi-fidelity benchmarks with first class support for user priors

https://automl.github.io/mf-prior-bench/

Apache License 2.0

8 stars 2 forks source link

Closed Sohambasu07 closed 5 months ago

Sohambasu07 commented 6 months ago

Changes made:

Clipped PD1 surrogates' results by metric's max bounds
Changed minimum fidelity values of MFHartmann, JAHSBench, PD1 imagenet and PD1 cifar100 to 1
Corrected surrogate names of PD1 cifar100 and PD1 translate_wmt
Changed jahs_bench.Benchmark(task=self.task_id) to (task=task)

Sohambasu07 commented 5 months ago

Looks good.

Has this been tested with a clean download?

Hey, yes. I've run all tests and tested that the surrogates return valid values as well