Tidy the PyTorch Docker build scripts and configuration, in order to:
Enable full logging from the docker build process - to simplify investigation should something go wrong
Read Docker credentials from Vault.
Limit the scope during which the Docker credentials are in play.
Push PyTorch Docker image to registry upon successful build.
Simplify Buildkite pipeline steps so that the QA tests are only run upon success of Linux x86_64 builds
Use a c2-standard-16 GCP machine for the Docker build step - this greatly improves build times.
Trigger the appex-qa-stateful-custom-ml-cpp-build-testing pipeline after successfully building the ml-cpp code in the newly built PyTorch Docker image - This may be subject to change, but does serve the purpose of some QA testing on the ml-cpp code.
Tidy the PyTorch Docker build scripts and configuration, in order to:
Linux x86_64 builds
c2-standard-16
GCP machine for the Docker build step - this greatly improves build times.appex-qa-stateful-custom-ml-cpp-build-testing
pipeline after successfully building theml-cpp
code in the newly built PyTorch Docker image - This may be subject to change, but does serve the purpose of some QA testing on theml-cpp
code.