-
Notifications
You must be signed in to change notification settings - Fork 25.9k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Add setup for TPU CI to run every hour. #6219
Conversation
From previous PR: LysandreJik 4 hours ago Member By the way, why is this model specific? Are we using these values somewhere? zcain117 19 minutes ago Author The local bertBasedCase is just a name of that jsonnet variable, it just needs to be a unique string but several punctuation marks are not allowed in variable names. We recommend 1 variable like this per test. On our team we have 1 file per "family" of tests, e.g. 1 file for pytorch resnet50 where the file contains [short test, long test] X [v2-8 TPU, v3-8 TPU, v100 GPU(s)] for the same model on the same ML framework. The modelName and frameworkPrefix are just used in generating the name of the GKE job. For example, a recent run was named: hf-bert-based-case-example-v3-8-vtgrj. Future runs will look like hf-bert-base-cased-example-v3-8-xxxxx |
Codecov Report
@@ Coverage Diff @@
## master #6219 +/- ##
==========================================
- Coverage 78.54% 78.44% -0.11%
==========================================
Files 148 146 -2
Lines 27196 26586 -610
==========================================
- Hits 21361 20855 -506
+ Misses 5835 5731 -104
Continue to review full report at Codecov.
|
@LysandreJik let me know if there's anything else you can think of here |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Re-organized the template a bit and set to run once per dey. Thanks a lot @zcain117!
Use GKE to run TPU CI testing once per hour using latest code in master branch.