If a third party writes buggy code and it involves a Keras model, that's not "a bug in Keras", for Christ's sake
Show this thread
-
So do, in fact, check it yourself, via a unit test -- it's never wasted time. For any codebase that's actively maintained, you should always aim for 100% coverage
Show this thread
Unit tests are the most useful maintenance tool there is -- they enable you to make changes with confidence.https://twitter.com/fchollet/status/1273713171111440385?s=19 …
-
-
Do you know of any useful examples of "good" unit testing in Keras that you don't mind sharing?
-
Well, you can check out the utility we use for standardized tests on all built-in layers -- it's not comprehensive but it covers the bits that are common to all layers. Search for layer_test
- Show replies
New conversation -
-
-
I usually accept a lot of your opinions. But I hope you might consider to reflect on this one: test are only guardrails, reasonability of your code will allow you to change code with confidence. Even with 100% test coverage does not mean 100% of input space coverage.
-
I still agree you should write test checking the properties of your custom layers, but testing is not equal to confidence, it is only a check.
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.