René Sass

Check this out: https://github.com/automl/SMAC3/blob/development-2.0/smac/callback.py Integrated the callbacks here: https://github.com/automl/SMAC3/blob/development-2.0/smac/smbo.py Docs is also (nearly) ready: https://automl.github.io/SMAC3/development-2.0/advanced_usage/callbacks.html

Hi Johannes, thank you *very* much for finding this out. We will discuss this internally and will change the behaviour in SMAC 2.0.

Hi Johannes, I tried out your fix but it results in not evaluating any trials anymore. Did you change more code to make it run?

Can you fork https://github.com/automl/SMAC3/pull/875 and check if your example still run as expected? Please use this "template" to get started: https://automl.github.io/SMAC3/development-2.0/examples/2_multi_fidelity/1_mlp_epochs.html#sphx-glr-examples-2-multi-fidelity-1-mlp-epochs-py

We decided to rewrite the intensifier completely in version 2.0.0a2. The new SH/HB will be much easier to understand without the need of the run tracker. The issue should then...

Would be super cool but could be kinda complicated when having hierarchical/conditional hyperparameters.

Since we drop the pcs format in v2.0 we could integrate this feature too. What do you think @mfeurer?

Would be nice.

That would be indeed beneficial. I will integrate this feature in our upcoming Version 2.0. I'm thinking of the following: - Scenario gets a new argument called `termination_cost_threshold` - SMAC...

I just implemented it: https://github.com/automl/SMAC3/pull/875/commits/c4823e0711db13fa87969159564eb766a4e29b8a