MITLibraries / tacos-detectors-lambdas
99%
main: 99%

LAST BUILD BRANCH: dependabot/pip/sentry-sdk-2.43.0
DEFAULT BRANCH: main
Repo Added
Files 2
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst

LAST BUILD ON BRANCH tco-157
branch: tco-157
CHANGE BRANCH
x

coverage: 99.219% (+0.05%) from 99.167%
15859894832

Pull #28

github

matt-bernhardt 
Respond to code review feedback

This makes a few changes that have come about while responding to code
review:

1. We change the return from the prediction to be a boolean, rather than a
string. There will be a related change to TACOS to deal with this.

2. We back away entirely from using check_is_fitted during the lambda's
operation. As Graham and I talked about the issues at play, and what the
proper response should be to an unfitted model, we realized that runtime is
not the concern - given that the model is being committed to version control,
the proper defensive point is during the PR, and as part of unit tests.

It turns out that several of our existing unit tests blow up when an unfitted
model is used in the codebase. One of those has been explicitly renamed as a
test against an unfitted model, although two other tests (the tests of the
positive and negative prediction outcomes) will also fail if the model is not
fitted.

Side effect:
The application now has something of an unprotected belly, which can be
exposed if we ever get to the point of placing the model in an s3 bucket
for loading at runtime rather than deploying after code review and unit
testing.

3. The Makefile gets changed because I'd reversed the slash when declaring
the expected content type. It worked locally, but failed when deployed to
AWS.

4. We also remove a deprecated linter from the ignore list in pyproject.toml
upon feedback from the linter.
Pull Request #28: Implement prediction logic via neural network algorithm

10 of 10 new or added lines in 1 file covered. (100.0%)

127 of 128 relevant lines covered (99.22%)

0.99 hits per line

Relevant lines Covered
128 RELEVANT LINES 127 COVERED LINES
0.99 HITS PER LINE
Source Files on tco-157

Recent builds

Builds Branch Commit Type Ran Committer Via Coverage
15859894832 tco-157 Respond to code review feedback This makes a few changes that have come about while responding to code review: 1. We change the return from the prediction to be a boolean, rather than a string. There will be a related change to TACOS to deal wit... Pull #28 matt-bernhardt github
99.22
15834273464 tco-157 Respond to code review feedback - This drops a try/catch block from the PredictHandler's load_model method, letting sklearn throw its own NotFittedError in the event that we deploy a model that isn't fitted. This should still be sent to Sentry, a... Pull #28 matt-bernhardt github
99.23
15833352285 tco-157 Implement real prediction logic (w/side effects) ** Why are these changes being introduced: The lambda needs to actually do work, not just always return true. ** Relevant ticket(s): * https://mitlibraries.atlassian.net/browse/tco-157 ** How d... Pull #28 matt-bernhardt github
99.26
15643573508 tco-157 Implement real prediction logic (w/side effects) ** Why are these changes being introduced: The lambda needs to actually do work, not just always return true. ** Relevant ticket(s): * https://mitlibraries.atlassian.net/browse/tco-157 ** How d... Pull #28 matt-bernhardt github
99.26
15639275756 tco-157 Implement real prediction logic (w/side effects) ** Why are these changes being introduced: The lambda needs to actually do work, not just always return true. ** Relevant ticket(s): * https://mitlibraries.atlassian.net/browse/tco-157 ** How d... Pull #28 matt-bernhardt github
99.26
15637570313 tco-157 Implement real prediction logic (w/side effects) ** Why are these changes being introduced: The lambda needs to actually do work, not just always return true. ** Relevant ticket(s): * https://mitlibraries.atlassian.net/browse/tco-157 ** How d... Pull #28 matt-bernhardt github
99.26
15621165493 tco-157 Test for unfitted model Pull #28 matt-bernhardt github
99.28
15619402119 tco-157 Implement real prediction logic This does a lot: - pyproject.toml gets updated because sklearn does have type hinting so was getting flagged by mypy - The neural network model has been retrained with lowercase parameter names, and with parameters... push matt-bernhardt github
96.4
15614481774 tco-157 Adds pandas-stubs push matt-bernhardt github
99.17
See All Builds (106)