yep

With this scenario, your data should be updated when running the pipeline

Hi DeliciousBluewhale87

Can you share the version you are using? Did you get any other logs? maybe from the pod?

https://hub.docker.com/r/nvidia/cuda/tags?page=1&ordering=last_updated&name=11.0-

👍

try with Cuda11.0 image,

Task B is a clone of Taks A. Does B store the information that it was cloned from A somewhere?

You can add any user properties you like to any task, so maybe “origin” : <task_id> will do the work?

Hi DeliciousBluewhale87

compute time : The maximum compute time in minutes (sum of execution time on all machines). When time limit is exceeded, all jobs aborted.
jobs : set the maximum number of experiments for the optimization.
time : set the maximum time limit for the optimization process

This means that you still have 94% of the total run time, 91% of the total iterations and 60% of the jobs in your HPO controller budget.

Hi DeliciousBluewhale87 ,

You can just get a local copy of the dataset with ds.get_local_copy() , this will download the dataset from the dataset task (using cache) and return a path to the downloaded files.

Now, in this path you’ll have all the files that you have in the dataset, you can go over the files in the dataset with ds.list_files() (or ds.list_files()[0] if you have only 1 file) and get the one you want

maybe something like:

` ds_path = ds.get_local_copy()
iri...

Hi DeliciousBluewhale87 ,

You can get the latest dataset by calling Dataset.get :

from clearml import Dataset ds = Dataset.get(dataset_project="dataset-project", dataset_name="dataset-task-name")This will return you the latest dataset from the project

you can just running with the env var at the beginning:

YOUR_ENV_VAR_NAME=/path/to/my/data clearml-agent …

Hi SmarmySeaurchin8 ,

You can configure each trains-agent to run with a different image. In your trains.conf file, under agent.default_docker.image section, specify the image you want the trains-agent to run with. When you have a value in this section, and with empty BASE DOCKER IMAGE, you can avoid changing it manually.

Can this do the trick?

Hi SmarmySeaurchin8 ,

The trains-agent default uses the ~trains.conf file for credentials, can you verify the api section in this file?

Hi SmarmySeaurchin8

You can configure TRAINS_CONFIG_FILE env var with the conf file you want to run it with. Can this do the trick?

You can configure env vars in your docker compose, but what is your scenario? Maybe there are some other solutions

For the trains-agent , you have an option to specify the trains.conf file you want it to run with. just start the trains-agent with trains-agent --config ~/trains_agent.conf (where ~/trains_agent.conf is your ~/trains.conf file for the agent run).

how could I configure this in the docker compose?

Do you mean to env vars?

Hi ThankfulOwl72 ,

You can create only one main execution Task. In the code you wrote, you are trying to have two tasks, which causing the exception. You can read more about the task object in the https://allegro.ai/docs/task.html#trains.task.Task .

The reuse_last_task_id will create a new task, which is not the default for https://allegro.ai/docs/task.html#task.Task.init (will override the last one)

What is your use case? maybe I can help with that

BTW, you can use ` Task.init...

https://clear.ml/docs/latest/docs/references/sdk/automation_controller_pipelinecontroller#pipelinedecoratorcomponent

which clearml version are you using? did you add it and re run it without clearml-agent?

Answered at https://clearml.slack.com/archives/CTK20V944/p1654190121555049?thread_ts=1654190039.012519&cid=CTK20V944 🙂

Hi OddShrimp85 , Do you get any errors from the server?

👍 can you try with secure as true ?

Hi OddShrimp85 ,

Whats the clearml version you are using? Do you have boto3 installed?

👍
this is a message about configuration sync.

Its allow you to change scheduler in runtime by editing the Task configuration object

Hi VexedCat68 ,

the scheduler print

Not sure what do you mean here, can you add an example?

yes -

task.upload_artifact('local json file', artifact_object="/path/to/json/file.json")

Hi VexedCat68

Is it possible to write text file and see it in results of the experiment?

You can upload any file as an artifact to your task, try:

task.upload_artifact(name="results", artifact_object="text_file.txt")

I want to use it to version data as in keeping a track of what images have been trained on. Or is there a better way of data versioning in ClearML?

You can use https://clear.ml/docs/latest/docs/clearml_data/ for making the data accessible from every machine...

Is it possible to write text file and see it in results of the experiment?
You can upload any file as an artifact to your task, try:

task.upload_artifact(name=“results”, artifact_object=“text_file.txt”)

Notice the max preview for an artifact is 65k, and it is suggested to have one file like this (and not for every iteration for example)

In the self-hosted we do not have user permissions, so every user sees all the data.

You can always run the command with CLEARML_CONFIG_FILE=<YOUR clearml.conf file path>
and then the running command

For the agent you can add --config-file with the path to the configuration file