Hi ObedientToad56 🙂

My question is on how the deployment would be once we have verified the endpoints are working in a local container.

I isn't the deployment just running the inference container? You just open up the endpoints towards where you wanna server, no?

@<1533619725983027200:profile|BattyHedgehong22> , it appears from the log that it is failing to clone the repository. You need to provide credentials in clearml.conf

the experiments themselves 🙂

Imagine if you have very large diffs or very large (several mb) configuration files logged into the system - this is sitting somewhere in the backend on a database

Can you compare the installed packages between the original experiment to the cloned one? Do you see anything special or different between the two?

What is the setup that they do the training on?

Because I think you need to map out the pip cache folder to the docker

Hi @<1536881167746207744:profile|EnormousGoose35> , support for the PRO version is given here on community slack channel.

Let me play with it a bit and see if I can find more 🙂

Hi GentleSwallow91 ,

1. When using jupyter notebooks its best to do task.close() - It will bring the same affect you're interested in
2. If you would like to upload to the server you need to add the following parameter to your Task.init() The parameter is output_uri. You can read more here - https://clear.ml/docs/latest/docs/references/sdk/task#taskinit
   You can either mark it as True or provide a path to a bucket. The simplest usage would be ` Task.init(..., output_uri...

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)

OK, there appears to be a github issue relating this:
https://github.com/allegroai/clearml/issues/388
I was right regarding encountering this before. People have asked for this feature and I think it appears to be a priority to add as a feature.

You can circumvent auto logging with the following:
task = Task.init(..., auto_connect_frameworks={'pytorch': False})However you will need to log other models manually now. More information is in the github issue 🙂

from clearml import Task

task = Task.init(,...)
print("hello world!")

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

Also, can you copy here the contents of your docker-compose file here?

JitteryCoyote63 , are you on a self hosted server? It seems that the issue was solved for 3.8 release and I think should be released to the next self hosted release

You can clone it via the UI, enqueue it to a queue that has a worker running against that queue. You should get a perfect 1:1 reproduction

Hi @<1774245260931633152:profile|GloriousGoldfish63> , this feature is waiting enablement on clearml-serving side and will be supported in the next release

Hi @<1577468638728818688:profile|DelightfulArcticwolf22> , it looks like you're trying to update some parameter of a task after it finished running...

The CLI

Hi @<1546665634195050496:profile|SolidGoose91> , when configuring a new autoscaler you can click on '+ Add item' under compute resources and this will allow you to have another resource that is listening to another queue.

You need to set up all the resources to listen to the appropriate queues to enable this allocation of jobs according to resources.

Also in general - I wouldn't suggest having multiple autoscalers/resources listen to the same queue. 1 resource per queue. A good way to mana...

So, I went to the link

in order to use it like Postman. Testing the API without using Python. It was ChatGPT that directed me there, and it is kind of a nice way to validate the API

I would ignore anything that ChatGPT says about ClearML (and most other things too)

This description in the

add_tags()

doc intrigues me tho, I would like to remove a tag from a dataset and add it to another version (eg: a

used_in_last_training

tag) and this method seems to only add new tags.

I see. Then I think you would need to do this via the API:
https://clear.ml/docs/latest/docs/references/api/tasks#post-tasksupdate
or
https://clear.ml/docs/latest/docs/references/api/tasks#post-tasksupdate_batch

I think you need to provide the app pass for github/butbucket instead of your personal password

Feels like a cookie issue to me

Hi @<1524922424720625664:profile|TartLeopard58> , projects and many other internals like tasks are all saved in internal databases of the ClearML server, specifically in mongo & elastic

Hi @<1578555761724755968:profile|GrievingKoala83> , there is no such capability in the open source. To add new users you need to edit the users file.

In the Scale/Enterprise licenses you have full user management including role based access controls

Does ClearML support running the experiments on any "serverless" environments

Can you please elaborate by what you mean "serverless"?

such that GPU resources are allocated on demand?

You can define various queues for resources according to whatever structure you want. Does that make sense?

Alternatively, is there a story for auto-scaling GPU machines based on experiments waiting in the queue and some policy?

Do you mean an autoscaler for AWS for example?

Hi @<1576381444509405184:profile|ManiacalLizard2> , it looks like the default setting is still false

I think it sits somewhere in the UI code

Can you please open a github issue so the issue can be followed?