Hi ReassuredTiger98
I think it used to be the default and then it was removed, it has no real affect on performance but it remove all asserts ... what is your use case ? do you see any performance gains ?

Yey!

My question is what should be the path to the requirements.txt file?
Is it relative to the repo base?

This is actually in runtime (i.e. when running the code), so relative to the working directory. Make sense ? (you can specify absolute path, probably something I would avoid in the code base though...)

Hi WickedGoat98
This sounds like a great design (obviously you have scale in mind 😉 ) Feel free to ask "stupid" questions, based on what you already wrote I doubt they will be
A few questions that come to mind (probably a few others after):
You mentioned FS synchronization, from where? i.e. what is the single source of truth ? K8s (Rancher 2.0 is basically k8s manager) can take care of mounting volumes, so no need to sync, is this a valid solution ?

BTW : (you can drag and drop an i...

Hi @<1691620877822595072:profile|FlutteringMouse14>

In the latest project I created, Hydra conf is not logged automatically.

Any chance the Task.init call is not on the main script (where the Hydra is) ?

YummyFish22 can you point to the huggingface example you are using?

The experiment finished completely this time again

With the RC version or the latest ?

Are you saying you had that odd script entry-point created by calling Task.init? (To clarify this is the problem)
Btw after you clone the experiment you can always manually edit both entry point and working dir, which based on what you said should be "script.py" and "folder"

So this is why 🙂
an agent can only run one Task at a time.
The HPO (being a Task on its own) should run on the "services" queue, where the agent can run multiple "cpu controller" Tasks like the HPO.
Make sense ?

ResponsiveCamel97
could you attach the full log?

My driver says "CUDA Version: 11.2" (I am not even sure this is correct, since I do not remember installing code in this machine, but idk) and there is no pytorch for 11.2, so maybe it fallbacks to cpu?

For some reason it detect CUDA 11.1 (I assume this is what you have installed, the driver CUDA version is the highest it will support not necessary what you have installed)

You can get a mutable copy of the entire dataset (original version), with get_mutable_copy() Then change the files on the returned directory, then create a new Dataset with the parent dataset as the original verison, then sync the folder.
You can also just update the specific file (without needing to download the entire original version)

Hi CurvedHedgehog15

User aborted: stopping task (3)

?

This means "someone" externally aborted the Task, in your case the HPO aborted it (the sophisticated HyperBand Bayesian optimization algorithms we use, both Optuna and HpBandster) will early stop experiments based on their performance and continue if they need later

CheerfulGorilla72

yes, IP-based access,

hmm so this is the main downside of using IP based server, the links (debug images, models, artifacts) store the full URL (e.g. http://IP:8081/ http://IP:8081/... ) This means if you switched IP they will no longer work. Any chance to fix the new server to the old IP?
(the other option is somehow edit the DB with the links, I guess doable but quite risky)

Hmm I see what you mean. It is on the roadmap (ETA the next version 0.17, 0.16 is due in a week or so) to add multiple models per Task so it is easier to see the connections in the UI. I'm assuming this will solve the problem?

WackyRabbit7 my apologies for the lack of background in my answer 🙂
Let me start from the top, one of the goal of the trains-agent is to reproduce the "original" execution environment. Once that is done, it will launch the code and monitor it. In order to reproduce the original execution environment, trains-agent will install all the needed python packages, pull the code, and apply the uncommitted changes.
If your entire environment is python based, then virtual-environment mode is proba...

MuddySquid7 you mean you are creating them with TB ? or are you uploading them as debug images ?
Specifically in the ClearML UI, do you have it under "plots" tab or "debug samples" tab ?

ngrok to connect to the remote server at the office?
That makes sense, I guess this is the equivalent of using a VPN, from that point onward clearml-session can directly access the remote machine, right?

Hi MinuteWalrus85
This is great question, and super important when training models. This is why we designed a whole system to manage datasets (including storage querying, balancing data, and caching). Unfortunately this is only available in the paid tier of Allegro... You are welcome to https://allegro.ai/enterprise/ the sales guys.
🙂

you can run md5 on the file as stored in the remote storage (nfs or s3)

s3 is implementation specific (i.e. minio weka wassaby etc, might not support it) and I'm actually not sure regrading nfs (I mean you can run it, but it actually means you are reading the data, that said, nfs by definition I'm assuming is relatively fast access)
wdyt?

SmallDeer34 No worries, I'm happy to hear the issue disappeared 🙂

That makes sense...
Basically in the open-source version the approach is everyone sees everything for maximum transparency (and also ease of use). I know there are access-roles in the paid tier and vault for exactly these types of things...

Where do you currently save them? and how do you pass them to the remote machine ?

Now I need to figure out how to export that task id

You can always look it up 🙂
How come you do not have it?

mean? Is it not possible that I call code that is somewhere else on my local computer and/or in my code base? That makes things a bit complicated if my current repository is not somehow available to the agent.

I guess you can ignore this argument for the sake of simple discussion. If you need access to extra files/functions, just make sure you point the repo argument to their repo, and the agent will make sure your code is running from the repo root, with all the repo files under i...

note

/home/npuser/.clearml/venvs-builds/3.7/task_repository/commons-imagery-models-py

is the correct pat

So how come it is failing?
Can you also print sys.path just to be sure ?

How are you spinning the agents ?

Add '/' , like you would with a file system.
Task.init(project_name='main_project/sub_project', task_name='test')

Because we are working with very big files, having them stored at multiple locations is something we try to avoid

Just so I better understand, is this for storing files as part of a dataset, or as debug samples ?
In other words can two diff processes create the exact same file (image) ?

, I can see the shape is

[136, 64, 80, 80]

. Is that correct?

Yes that's correct. In case of the name, just try input__0
Notice you also need to convert it to torchscript

OutrageousSheep60

I found the task in the UI -

and in the

UNCOMMITTED CHANGES

execution section there is

No changes logged

This is the issue.

and then run the

session

via docker

clearml-session --docker nvidia/cuda:10.1-cudnn7-runtime-ubuntu18.04 \ --packages "clearml" "tensorflow>=2.2" "keras" \ --queue MY_QUEUE \ --verboseAre you running the "cleamrl-session" from your machine? (i.e. not from inside a docker) ?...