Aaron Klein comments

Results 32 comments of


                                            Aaron Klein

clean up epmgp.py

sure I can do that

Which version is the recommended version to use?

generally I recommend to use the current master branch

BUG: printed_every_n_steps and keep_every cannot be equal

good catch thanks, could you send a PR to fix it?

nose.proxy.UnsupportedOperation: fileno

Hi, it seems that this is the same problem as in #75 . Could you pull the master branch again?

the incompatibility between george.kernels.Matern52Kernel and george.kernels.TaskKernel in robo.fmin.mtbo

Unfortunately, the Taskkernel doesn't work with the latest george version anymore (automl/george development branch) and I don't know how to fix that with their new interface for kernels

Accessing optimizer's internal state

Hi, good point. You can already do that by instantiating the robo.solver.BayesianOptimization class and call the choose_next() function: For example: ``` bo = robo.solver.BayesianOptimization(model, acquisition, lower, upper, objective,..) X=None Y=None...

Accessing optimizer's internal state

Unfortunately yes. We are right now working on a new package that is more flexible than RoBO but contains the same functionality. We are planning it to release it in...

Accessing optimizer's internal state

yes it is online (https://github.com/amzn/emukit) but not yet officially released and under heavy development

ModuleNotFoundError: No module named 'sagemaker.interactive_apps'

Hey @austinmw sorry for the delay. Could you post a script to reproduce the error? From the error script it seems that you resume a previous tuning job? Does the...

ModuleNotFoundError: No module named 'sagemaker.interactive_apps'

I assume this issue is solved. Feel free to re-open