I think these env varibles might be relevant to you:
CLEARML_AGENT_SKIP_PIP_VENV_INSTALL
CLEARML_AGENT_SKIP_PYTHON_ENV_INSTALL
https://clear.ml/docs/latest/docs/clearml_agent/clearml_agent_env_var

I think this is the relevant one:
https://stackoverflow.com/questions/50305112/pip-install-timeout-issue

You can add it to your pip configuration so it will always be taken into account

Hi @<1671689442621919232:profile|ItchyDuck87> , did you manage to register directly via the SDK?

Hi @<1734020162731905024:profile|RattyBluewhale45> , what version of pytorch are you specifying?

CUDA is the driver itself. The agent doesn't install CUDA but installs a compatible torch assuming that CUDA is properly installed.

I suggest running it in docker mode with a docker image that already has cuda installed

OK, then just try the docker image I suggested 🙂

unrelated to the agent itself

Hi @<1787653566505160704:profile|EnchantingOctopus35> , I don't think you can clean up parents without damaging the children since they rely on that. I would suggest taking any data you don't want scrubbed and then creating a new version with it. Then delete the unrelated older datasets. What do you think?

Hi @<1524560082761682944:profile|MammothParrot39> , I think you need to run the pipeline at least once (at least the first step should start) for it to "catch" the configs. I suggest you run once with pipe.start_locally(run_pipeline_steps_locally=True)

Hi @<1636537816684957696:profile|CooperativeGoat65> , you can change the api.files_server section of the configuration file to point to your s3 bucket

ExuberantParrot61 , I'm not sure I understand the entire setup. can you please elaborate?

ReassuredTiger98 , I played with it myself a little bit - It looks like this happens for me when an experiment is running and reporting images and changing metric does the trick - i.e reproduces it. Maybe open a github issue to follow this 🙂 ?

Hi @<1734744933908090880:profile|WorriedShells95> , I suggest going through the documentation - None

Hi @<1625303791509180416:profile|ExasperatedGoldfish33> , I would suggest trying pipelines from decorators. This way you can have very easy access to the code.
None

Hi @<1673501397007470592:profile|RelievedDuck3> , no you don't. The basics can be run with a docker compose 🙂

Tell me if it helps 🙂

I think the 3rd one, let me know what worked for you

Hi @<1570220844972511232:profile|ObnoxiousBluewhale25> , I think the API server can delete things only from the files server currently. However the SDK certain has the capability to delete remote files

I think you can set this code wise as well - https://clear.ml/docs/latest/docs/references/sdk/task#taskforce_requirements_env_freeze

Hi MistakenDragonfly51 , regarding your questions:
ClearML has a model repository built in. You can load an input model using InputModel module ( https://clear.ml/docs/latest/docs/references/sdk/model_inputmodel ). Also, you can fetch the models of an experiment using Task.get_models() - https://clear.ml/docs/latest/docs/references/sdk/task#get_models Can you elaborate on how this config looks in the UI when you view it?

GiganticTurtle0 , does it pose some sort of problem? What version are you using?

Hi @<1566596960691949568:profile|UpsetWalrus59> , I think this basically means you have an existing model and it's using it as the starting point.

JitteryCoyote63 , reproduces on my side as well 🙂

AbruptWorm50 , can you confirm it works for you as well?

Looks like you're having issues connecting to the server through the SDK. Are you able to access the webUI? Is it a self hosted server?

The project should have a system tag called 'hidden'. If you remove the tag via the API ( None ) that should solve the issue.

How was the project turned to hidden?

Are you seeing any errors in the webserver container?

I think this is because you're working on a "local" dataset. Only after finalizing the dataset closes up. Can you describe your scenario and what was your expected behavior?