Alright. Thanks CostlyOstrich36

Yea. Added an issue. We can follow up from there. Really hope that clearml serving can work, is a nice project.

And just a suggestion which maybe I can post in GitHub issue too.
It is not very clear what are the purpose of the project name and name, even after I read the --help. Perhaps this is something that can be made clearer when updating the docu?

Not exactly sure yet but I would think user tag for deployed make sense as it should be a deliberated user action. And additional system state is required too since a deployed state should have some pre-requitise system state.

I would also like to ask if clearml has different states for a task, model, or even different task types? Right now I dun see differences, is this a deliberated design?

U want to share your clearml.conf here?

@<1523701070390366208:profile|CostlyOstrich36> Yes. I'm running on k8s

Thanks AgitatedDove14 . Specifically, I wanted to use my own clearml server and Triton. Thus, I attempted to use --engine-container-args during launch but error saying no such flag. Looked into --help but I guessed it is not updated yet.

I was browsing clearml agent gihub and saw this. Isn't this for spinning up clearml-agent in a docker and perform like a daemon?

@<1523701205467926528:profile|AgitatedDove14> I still trying to figure out how to do so. Coz when I add a task in queue, clearml agent basically creates a pod with the container. How can I make a task that does a helm install or kubectl create deployment.yaml?

To clarify, there might be cases where we get helm chart /k8s manifests to deploy a inference services. A black box to us.

Users may need to deploy this service where needed to test out against other software components. This needs gpu resources which a queue system will allow them to queue up and eventually get this deployed instead of hard resource allocation to this purpose

It return false. Just to share abit more, I have the requirements.txt in gitlab with my codes and are in folders. Do I need to provide a gitlab path?

Ok. Can I check that only the main script was stored in the task but not the dependent packages?

I guess the more correct way is to upload to some repo where the remote task can still pull from it?

@<1523701205467926528:profile|AgitatedDove14> do u mean not using helm but fill up the values and install with the yaml files directly? E.g. kubectl apply ...

Thanks

Hi ExasperatedCrab78 I managed to get it. It was due to ip address set in examples.env.

@<1526734383564722176:profile|BoredBat47> Just to check if u need to do update-ca-certificates or equivalent?

CostlyOstrich36 I mean the dataset object in clearml as well as the data that is tied to this object.

The intent is to bring over to another clearlml setup and keep some form of traceability.

Nice. It is actually dataset.id .

@<1523701070390366208:profile|CostlyOstrich36> This is output_uri or where do I put this url?

Thanks @<1523701205467926528:profile|AgitatedDove14> . what I could think of is to write a task that may run python subproecss to do "helm install". In those python script, we could point to /download the helm chart from somewhere (e.g. nfs, s3).

Does this sound right to u?
Anything that I was wondering is if we could pass the helm charts /files when we uses clearml sdk, so we could minimise the step to push them to the nfs/s3.

@<1523701205467926528:profile|AgitatedDove14> I looking at a queue system which clearml q offers that allow user to queue job to deploy an app / inference service. This cam be as simple as a pod or a more complete helm chart.

Can clearml-serving does helm install or upgrade? We have cases where the ml models do not come from the ml experiments in clearml. But would like to tap on clearml q to enable resource queuing.

Hi @<1523701070390366208:profile|CostlyOstrich36> , basically

• I uploaded dataset using clearml Datasets. The output_uri is pointed to my s3, thus the dataset is stored in s3. My s3 is setup with http only.
• When I retrieve the dataset for training, using Dataset.get() , I encountered ssl cert error as the url to retrieve data was https://<s3url>/... instead of s3://<s3url>/... which is http. This is weird as the dataset url is without https.
• I am not too sure why and I susp...

Example i build my docker image using a image in docker hub. In this image, i installed torch and cupy packages. But when i run my experiment in this image, the packages are not found.

Yes, I ran the experiment inside.

I guess we need to understand the purpose of the various states. So far only "archive, draft, publish". Did I miss any?

Ah I think I was not very clear on my requirement. I was looking at porting project level, not entire clearml data over. Is it possible instead?

I see. Was wondering any advantage to do it any of the ways.

Do u have an example of how I can define the packages to be installed for every steps of the pipeline?

Hi Bart, yes. Running with inference container.

I figured out that it maybe possible to do these
experiment_task = Task.current_task()
OutputModel(experiment_task ).update_weights(' http://model.pt ') to attach it to the ClearML experiment task.