Replies: 2 comments
-
Sounds good, one question: Will this be more like a benchmarking or unit testing? For benchmarking, we probably want to do periodic run on GCP instead of pytest. Or maybe we want both? |
Beta Was this translation helpful? Give feedback.
0 replies
-
My initial plan was just to test that the files are available and have the right permissions. This only needs to be done on occasion. Would the same be true for the benchmarks or do you want to run them manually @chenmoneygithub? |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
We are adding pretrained tokenizers (PR #287) and models (PR #299) to the package in recent PRs. While we will want to periodically verify the the necessary files are available and have the right permissions, we cannot create unit tests with large network calls (especially when models are as much as 1GB).
I propose to add an
network_tests/
or similarly named folder outside thekeras_nlp
package folder where we can add tests for loading pretrained objects that can be run manually as needed withpytest network_tests/
.Beta Was this translation helpful? Give feedback.
All reactions