fix up indexing of summed_mask in shapesys, staterror

[kratsg]

Description

The main difficulty with shapesys/staterror as it affects multiple (non-consecutive) bins is the way we do masking. For situations where we need to figure out whether to enable the mask or not, it was filling in -1 when the parameter index was 0 which is not good. Additionally, tensorflow.gather() is not happy with -1.

Checklist Before Requesting Reviewer

• Tests are passing
• "WIP" removed from the title of the pull request
• Selected an Assignee for the PR to be responsible for the log summary

Before Merging

For the PR Assignees:

• Summarize commit messages into a comprehensive review of the PR

[coveralls]

Coverage increased (+0.003%) to 97.714% when pulling eb6788f on fix/indexingCombinedMods into ea52590 on master.

[coveralls]

Coverage increased (+0.003%) to 97.714% when pulling eb6788f on fix/indexingCombinedMods into ea52590 on master.

[kratsg]

@matthewfeickert done ✅

[matthewfeickert]

LGTM!

[lukasheinrich]

Hi, the -1 was the right index in order to pick up the default value which was appended at the end of the pars in apply, do i‘m not sure this is the right change

(Sorry, On phone)

[kratsg]

Hi, the -1 was the right index in order to pick up the default value which was appended at the end of the pars in apply, do i‘m not sure this is the right change

This broke in tensorflow. gather(....) kept crashing since it can't access the index. Do you want to change -1 to len(inds) - 1?

[kratsg]

We also additionally changed shapesys in refactor/megachannel the same way to fix it so I just applied the same thing there to here.

[lukasheinrich]

TF can‘t use negative indices?

[kratsg]

Nope. It's gather() complains about the -1.

[kratsg]

tensorflow/tensorflow#11101

[lukasheinrich]

Yeah then I guess it should be len-1