Hi @<1523701205467926528:profile|AgitatedDove14>
I'm having a similar issue.

Also notice the cleaml-agent will not change the entry point of the docker meaning if the entry point does not end with plain bash, it will not actually run anything

Not sure I understand how to run a docker_bash_setup_script and then run a python script - Do you have an example? I could not find one.
Here is our CLI command

clearml-task --name <TASK NAME> \
		--project <PRJ NAME> \
		--repo git@github.com:<REPO NAME>.git \
		--script pytorch/main.py \
		--args dataset_dir='.'  \
		--task-type	training 	\
		--docker_bash_setup_script pre_installations.sh \
		--docker gcr.io/path/to/image \
		--queue default

running on clearml-agent Version 1.9.0