search

IBM / mi-prometheus

Enabling reproducible Machine Learning research
http://mi-prometheus.rtfd.io/
Apache License 2.0
42 stars 18 forks source link

Standardize params names across methods, problems etc. #73

Open sesevgen opened 5 years ago

sesevgen commented 5 years ago

There is inconsistent naming across params. For example, directories are sometimes called 'dir' and sometimes called 'folder'. We should perhaps decide on a unified set of standard names for source and target folders etc, and then change init and configs to reflect it.

tkornuta-ibm commented 5 years ago

I totally agree, I already started that with CLEVR, Sort-of-CLEVR, MNIST and CIFAR10.

I used "data_folder", with default value set to "~/data/mnist" (in the case of MNIST;) )

Besides, along with the introduction of helpers that can operate directly on problems/models, we have made a decision that all problems/models should provide default values for all parameters, enabling to instantiate them without configuration file.

Please refer to: https://github.com/IBM/mi-prometheus/blob/194cedec3431c7b56d1c1b2fed1a6e75e9fec265/miprometheus/problems/image_to_class/mnist.py#L91

@sesevgen Please keep that in mind now, probably I should create a separate issues for this...

vmarois commented 5 years ago

Agree on this as well. Alongside unifying the names used for the 'common' arguments, we should also make more explicit the several arguments that a Problem / Model class expects. This is indicated in #34 🙂