Open Christovis opened 5 years ago
As far as I understood Sacred now, it is too restrictive, a thing that is discussed in detail also here. What I mean by 'restrictive' is: a) Experiment settings have to be defined in a static way within the python script containing the experiment, e.g. it is not possible to do it in a dynamical way and have one python script loading the configuration for different ML-models without loosing sacrets benefits. b) is there another option to avoid using https://mlab.com/ ? c) ...
Rewrite models (sklearn and pytorch, order according to priority) to utilize sacred and tensorboard for management of experiments.