Support tensorflow 2.16 #102
Conversation
| "gpflow>=2.6.3", | ||
| "numpy", | ||
| "gpflow>=2.9.2", | ||
| "numpy<2", |
There was a problem hiding this comment.
What is the issue with numpy>=2?
There was a problem hiding this comment.
It's a major, breaking release: e.g. see https://numpy.org/devdocs/numpy_2_0_migration_guide.html.
We'll certainly want to have a conversation about supporting it at some point (possibly company wide), but at this point it probably makes sense to just wait until the dust settles (it was release just a few days ago).
| @@ -20,9 +20,10 @@ jobs: | |||
| check-and-test: | |||
| runs-on: ubuntu-20.04 | |||
| strategy: | |||
| fail-fast: false | |||
There was a problem hiding this comment.
Just to check, presumably this wouldn't make testing much slower anyway, as gpflux tests are very fast; correct?
There was a problem hiding this comment.
The tests take just as long as before if they all pass. However, if one of them fails early on, then previously we would have cancelled all the others, while now we will wait until they all finish (around 10 minutes).
Support and test against TF 2.16.
Note that this also sets fail-fast to false, which makes handling fragile tests easier.