He/Him 🙂

Would adding support for some sort of post task script help? Is something already there?

Optuna

Do we support GPUs in a) docker mode b) k8s glue?

Thanks AgitatedDove14 - i get overall what you are saying. Have to get glue setup, which I couldn’t understand fully, so that’s a different topic 🙂

no task.close()

The job itself doesn’t have any other param

how do you see things being used as the most normal way?

🙏

The Optimizer task is taking a lot of time to complete. Is it doing something here:

You mean the job with the exact same arguments ?

Yes

No, all of them completed!

Beyond this have the UI running, have to start playing with it. Any suggestions for agents with k8s?

https://github.com/allegroai/clearml-agent/blob/aede6f4bac71c8fc56e7cf982318a48527953a3c/clearml_agent/glue/k8s.py#L217

Thanks for the fast responses as usual AgitatedDove14 🙂

AgitatedDove14 - does having this template work for updating hte base image:

` spec:
containers:

• image: nvidia/cuda:11.4.1-cudnn8-runtime-ubuntu20.04 `

I am seeing that it still picks up nvidia/cuda:10.1-cudnn7-runtime-ubuntu18.04

AgitatedDove14 aws autoscaler is not k8s native right? That's sort of the loose point I am coming at.

AgitatedDove14 - tried exit(0) from the notebook and it worked

Having a pipeline controller and running actually seems to work as long as i have them as separate notebooks

PipelineController with 1 task. That 1 task passed but the pipeline says running

Basic question - i am running clearml agent in a ubuntu ec2 machine. Does it use docker by default? I thought it uses docker only if I add the --docker flag?

That’s about what i do

I then install this wrapper SDK in my containers, notebook instances etc

What happens if I do blah/dataset_url ?

Can makes some PRs around this. Playing with some changes already for my setup

` if project_name is None and Task.current_task() is not None:
project_name = Task.current_task().get_project_name()

if project_name is None and not Task.running_locally():
task = Task.init()
project_name = task.get_project_name() `

Does a pipeline step behave differently?

Maybe two thing here:

If Task.init() is called in an already running task, don’t reset auto_connect_frameworks? (if i am understanding the behaviour right) Option to disable these in the clearml.conf

Any reason to not have those as two datasets?