RoughTiger69 , regarding the dataset loading, we are actually thinking of adding it as another "hyper parameter" section, and I think the idea came up a few times in the last month, so we should definitely do that. The question is how do we support multiple entries (i.e. two datasets loaded)? Should we force users to "name" the dataset when they "get it" ?

Regrading cloning, we had a lot of internal discussions on it, "Parent" is a field on a Task, so the information can be easily stored, th...

Can you add a full log from startup of both Elastic and apiserver containers?

Hi @<1547028074090991616:profile|ShaggySwan64> , how are you currently saving models? What framework are you using? Usually the output models are listed in the 'artifacts' section of a task and on the model side, there is the 'lineage' tab to see which task created the model and what other tasks are using it as input.

Hi @<1689446563463565312:profile|SmallTurkey79> , when this happens, do you see anything in the API server logs? How is the agent running, on top of K8s or bare metal? Docker mode or venv?

Are you running a self hosted server?

Hi @<1546303277010784256:profile|LivelyBadger26> , can you provide a snippet that reproduces this?

They do look identical, I think the same issue (If it's an issue) also affects https://clear.ml/docs/latest/docs/references/sdk/dataset/#list_added_files

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

Is this all happening when you're running locally? How many gpu's do you have/try to run on? Also, can you provide an example code snippet to try and run something basic to get a similar failure. I think I have a machine with multiple gpus that I can try playing on 🙂

Also, you need to restart the agent between changes in the config

Hi @<1547028074090991616:profile|ShaggySwan64> , You can try this. However, Elastic takes space according to the amount of metrics you're saving. Clearing some older experiments would free up space. What do you think?

A workaround can be to set up a local Minio server or upload to s3 directly, this way there shouldn't be a limit

Unless you're running in docker mode, then I think the task will continue running inside the container. Might need to check it

You can read up on the caching options in your ~/clearml.conf
You can make virtualenv creation a bit faster

PanickyMoth78 , let me check on that 🙂

Hi @<1523701260895653888:profile|QuaintJellyfish58> , if you run in docker mode you can easily add environment variables.

Can you elaborate a bit on your use case? If it's python code, why not just put it in the original file or import from the repo?

From the environment variable

Is output_uri defined for both steps? Just making sure.

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...

RotundSquirrel78 , you can go to localhost:8080/version.json

DilapidatedDucks58 , regarding internal workings - MongoDB - all experiment objects are saved there. Elastic - Console logs, debug samples, scalars all is saved there. Redis - some stuff regarding agents I think

Hi ShallowGoldfish8 ,

I'm not sure I understand the scenario. Can you please elaborate? In the end the model object is there so you can easily fetch the raw data and track it.

Hi @<1835488771542355968:profile|PerplexedShells66> , you can set that up directly with set_repo - None

Hi @<1695969549783928832:profile|ObedientTurkey46> , this capability is only covered in the Hyperdatasets feature. There you can both chunk and query specific metadata.
None

You need to authenticate your communication with the ClearML server initially somehow, no? Otherwise basically anyone can create credentials on your server...
After you have authentication you can create credentials via the terminal.

You can create up to 10 sets of credentials per user so if you plan on creating new credentials every time you want to run a job - this is an incorrect approach.

Therefor - you should create the credentials once and then use them as environment variables as you...

It's totally possible, I think you need to do research on it. There are probably a few ways to do it too. I see CLEARML_API_ACCESS_KEY & CLEARML_API_SECRET_KEY in the docker compose - None

You should do some more digging around. One option is to see how you can generate a key/secret pair and inject them via your script into mongoDB where the credentials are stored. Another way is to see how the UI ...

I think you would need to contact the sales department for this 🙂
None

Yes please 🙂

Hi @<1623491856241266688:profile|TenseCrab59> , can you elaborate on what do you mean spending this compute on other hprams? I think you could in theory check if a previous artifact file is located then you could also change the parameters & task name from within the code

Is it a self hosted server?