Hi all,
I have a use-case for the TFJob I was wondering if you could maybe help me out with.
to run my TF training, I have a process that builds a docker image that contains all dependencies and the entrypoint runs a binary that runs a training service
I want to run this in a TFJob so that I can run this service across the chief + workers with TF_CONFIG set properly. I tried by doing:
However when the chief and worker containers start they fail as the image I supply 1) doesn't contain flyte-execute 2) doesnt have the correct entrypoint.
What is the proper way to achieve the desired behaviour?