Hi John, thank you. However, I could not find a hint there how to versionize tablular data. Our data is essentially a huge data frame where each ground truth data point is a row with a unique id. How can I track in clearML that this and that row was part of experiment x because it belonged to test/training data set y?

Hi John. If I start above mentioned script, I see 4 plots.

what else do you need? I am on client side, here but can find out more about our server installation, if that is your point

I am using clearML 1.13.2

Hi Martin, many thanks! that looks promising, will check

Hi Eugen, very helpful, thank you!

will check with our team, thank you, John

WebApp: 1.10.1-359 • Server: 1.10.1-359 • API: 2.24

this is the pipeline module until step1

I will add some code here soon.

here comes step 1:

I also want the subsequent tasks to access artefacts from the pipeline task (the config.yaml) as well as artefacts from prior tasks. Here and then it fails. not always

very strange. Sorry for putting in so much information

def dataset_versioning(
pipe_task_id: str
):

import os
from clearml import Dataset, Task, TaskTypes

from src.main.python.pkautopilot.configuration import PipelineConfig
from src.main.python.pkautopilot.constants import (TRAIN_FILE_NAME, TEST_FILE_NAME, CLEARML_BUCKET,
                                                   SUBDIR_DATASETS, CLEARML_PIPELINE_VERSION_DATASET, CLEARML_DATASET_KEY,
                                                   AWS_ACCESS_KEY, SUBD...

import os
from datetime import datetime
from typing import List, Optional, Sequence, Union
import json
import pickle

import typer
from clearml import Dataset, Task, TaskTypes
from clearml.automation import PipelineController
from clearml.utilities.config import config_dict_to_text

from src.main.python.pkautopilot.configuration import PipelineConfig
from src.main.python.pkautopilot.constants import *
from task1_dataset_versioning import dataset_versioning
from task2_hpo import hpo
from task3...

here is another failure which sometimes occurs along the pipeline: Launching step [model_training]
Launching step: model_training
Parameters:
{'kwargs/pipe_task_id': '6818423a08684b0d9972c92e5e38d6bc', 'kwargs_artifacts/dataset_id': '${version_dataset.id}.dataset_id', 'kwargs_artifacts/hpo_task_id': '${hyperparameter_optimization.id}.task_id'}
Configurations:
{}
Overrides:
{}
ClearML results page: [None](http://app.clearml.dev0.xrnd-modeling.dfman.info/projects/2a741ad0a93e49e48d5ecde83286e4...

Hi John. Thanks for your quick response. I looked carefully in the documentation and my use case seems to be standard for clearML. However I get this error message. clearml.Task - ERROR - Failed reloading task

I can also bring attachments here, if it make sense