PyTorch migration: Finishing touches #187
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
The previous explanation was taken from FATE and does not correctly describe the FETA approach.
timokau
force-pushed
the
migration-updates
branch
from
e4364f2 to
31a4a4e
Compare
7 tasks
kiudee
reviewed
Apr 8, 2021
7 tasks
kiudee
reviewed
Apr 8, 2021
kiudee
requested changes
Apr 8, 2021
The documentation of the scoring approach is now consistently delegated to the docstring of the scoring module. I think that is better, even if it is a bit harder for the user to find the descripütion. Duplicate documentation would just get out of sync and increases the friction for improvements. The class references should make it easy enough to find the relevant documentation.
This is somewhat contrary to the comment I wrote in the general choice test. I have since noticed a similar issue with RankNet in the tests. It seems that SELU just works better, at least for these tiny toy problems. Its already the default for the other pytorch based estimators. I think it makes sense to use it as the default for CmpNet as well.
It can be useful to compare the network sizes that are used for the different estimators at a glance. The CmpNet test specifications already list the `n_hidden` and `n_unit` arguments even though they match the default values. This change makes the specifications more consistent.
I have added them in alphabetical order, which is why the order has changed in `discretechoice.rst`.
timokau
force-pushed
the
migration-updates
branch
from
a293d1b to
16a4327
Compare
|
I fixed the typo, added cross-referencing, fixed the The references are not rendered as links. Sphinx gives some unrelated warnings, but does not complain about the references. It might be a configuration option. The sphinx warnings are related to the docstring indentation of some of the older estimators. I think the they could be fixed by using the latest version of The full access path should still make it easier to find the relevant documentation. |
kiudee
approved these changes
Apr 8, 2021
|
Thank you for the review :) |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description
This collects some misc fixes, changes and updates regarding the pytorch migration. I have updated the status in the README. The pytorch version does not have feature parity, but it is now at a point where it should be a viable replacement for at least some of the use cases.
Motivation and Context
Part of the ongoing pytorch migration.
How Has This Been Tested?
I ran the test suite and used our set of linters.
Does this close/impact existing issues?
Related to #125. I think the issue can be closed once the pytorch migration branch has been merged into master.
Types of changes
This contains various documentation changes as well as a breaking change to the CmpNet estimators.
Checklist: