Open eddyxu opened 2 years ago
Current GHA takes about 15 minutes to finish all tests. For rikai-scale project, this is not necessary.
Let's do some research to see where can we accelerated. A few potential low hanging fruit.
jdk8
scala 12
jdk 11
scala 13
mlflow
spark
For pretrained torchvision models, the pt files are saved $HOME/.cache/torch/hub/checkpoints. Just cache it in github action would save a lot networking time.
$HOME/.cache/torch/hub/checkpoints
Current GHA takes about 15 minutes to finish all tests. For rikai-scale project, this is not necessary.
Let's do some research to see where can we accelerated. A few potential low hanging fruit.
jdk8
andscala 12
against all Python version.jdk 11
+scala 13
against the latest supported python version.mlflow
andspark
docker images to speed up Python CI task.