FierceRabbit20 it seems the Pipeline Task that was created is missing the "installed requirements" section. How are you creating the actual pipeline Task? is this from code?

However for the task only cython is being installed for some reason.

Btw, to specify a custom package, add the path to that package to your requirements.txt (the path can also be a github link for example).

As I can see from the Pipeline's log, requirements.txt in being installed only for the master job (controller)

SmugDolphin23 btw, do you know how to pin the installation of some specific clearml verison inside clearml-agent ? (I want to use my custom version of clearml )

[package_manager.force_repo_requirements_txt=true] Skipping requirements, using repository "requirements.txt"
Try adding clearml to the requirements

If i change package_manager.force_repo_requirements_txt to false then clearml is successfully installed inside clearml-agent

it changed nothing

SmugDolphin23 , please see my requirements.txt -- it contains clearml==1.8.3.

From the Task's log (please, pay attention, not pipeline's log) you may see that none of these packages are installed.

requirements.txt:
attrs==22.2.0 # via # clearml # jsonschema certifi==2022.12.7 # via requests charset-normalizer==2.1.1 # via requests clearml==1.8.3 # via -r requirements.in click==8.1.3 # via typer colorama==0.4.6 # via typer commonmark==0.9.1 # via rich furl==2.1.3 # via clearml idna==3.4 # via requests importlib-resources==5.10.1 # via jsonschema jsonschema==4.17.3 # via clearml loguru==0.6.0 # via -r requirements.in numpy==1.24.0 # via clearml orderedmultidict==1.0.1 # via furl pathlib2==2.3.7.post1 # via clearml pillow==9.3.0 # via clearml pkgutil-resolve-name==1.3.10 # via jsonschema psutil==5.9.4 # via clearml pygments==2.13.0 # via rich pyjwt==2.4.0 # via clearml pyparsing==3.0.9 # via clearml pyrsistent==0.19.2 # via jsonschema python-dateutil==2.8.2 # via clearml pyyaml==6.0 # via clearml requests==2.28.1 # via clearml returns==0.19.0 # via -r requirements.in rich==12.6.0 # via typer shellingham==1.5.0 # via typer six==1.16.0 # via # clearml # furl # orderedmultidict # pathlib2 # python-dateutil tenacity==8.1.0 # via -r requirements.in typer[all]==0.7.0 # via -r requirements.in typing-extensions==4.4.0 # via # returns # rich urllib3==1.26.13 # via # clearml # requests zipp==3.11.0 # via importlib-resources

SuccessfulKoala55 Hello, thanks for the reply. Yes, that makes sense.

BTW, I also tried to specify the parameter packages of add_function_step . However, these packages were also ignored and I did not even get a warning message.

FierceRabbit20 I think the issue is that the controller code is not from source control, but a single file, and so there's no repo attached (hence the agent does not use the requirements.txt)

can you try setting the repo when calling add_function_step ?