How do I generate output from ML extractor trainer activity? For context, I am using classic DU (AI Center). In my extractor trainer activity, I am specifying the local output folder as well as the dataset in AI Center where I want the validated results to be placed but I do not see the results in either location. What am I doing wrong?
Hey @lomasd12, Make sure you’ve enabled “Export Results” in the Machine Learning Extractor Trainer activity properties. Without that, the output won’t be saved locally or sent to AI Center. Also, double-check that the in_Config("ExportsFolder") path is correctly initialized and accessible at runtime. For AI Center output, ensure that the API key, endpoint, and dataset permissions are correctly set and that the dataset exists under the selected project in AI Center. Finally, verify if any errors appear in the Output or Logs panel during execution.
Hello Jasim,
I do not see any properties in the trainer activity for enabling “export results.” Also, I do not see any permission configurations for the dataset in AI Center. Did I miss something? Aside from that, I have verified everything else to my knowledge
To enable result export in the ML Extractor Trainer activity, ensure the OutputFolder property is properly set (manually or via config). There’s no explicit “Export Results” toggle—the output works if the dataset and project are configured correctly. Use either a public or private dataset—not both—based on your setup. For private datasets, select the correct Dataset and Project, and avoid setting the API key or endpoint. If using a public dataset, provide only the API key and endpoint. Also, confirm your AI Center permissions (ML Package Contributor or Consumer) are sufficient. Make sure “Make dataset Public” is unchecked if working with a private dataset. Once these settings are in place, results will be saved to the specified folder automatically when you run the workflow.
Hey @lomasd12,
If the issue has been resolved then please mark as solution or let me know the current status.
Still no success
I have verified the API key and endpoint are supplied to the activity. The local output folder is correctly configured.
The dataset is configured as public
And I have full privilege in AI Center
Finally, my version of Ai Center/Automation suite is 2023.10
Hey @lomasd12,
You’ve configured everything correctly—API key, endpoint, and public dataset with full AI Center permissions.
However, one key point to check:
Make sure the dataset is already linked to the ML skill used in your “Train Extractors” activity.
If it’s not linked:
- Go to AI Center > ML Skills.
- Edit the skill and confirm your dataset is selected.
- Re-deploy if needed.
Still fails? Try regenerating and updating the API key.
So, a couple things. I do not see an option to link a dataset to an ML skill within the ML skill configuration screen. I think you may have intended for me to check the link between the dataset and the pipeline. Below is an image of my pipeline configuration. The dataset I am pointing to for training as well as evaluation is a folder within the dataset (CentralLincoln dataset) which is the same dataset specified in my train extractor activity. Are you suggesting this configuration may be wrong?
Hey @lomasd12,
You’re right — datasets are linked to pipelines, not ML Skills. Make sure the exact same folder path is used in both the pipeline and the Train Extractors activity. If your pipeline uses a subfolder (e.g., CentralLincoln/TrainingData), your activity must use that too. A mismatch in dataset paths will cause training issues.
The dataset name value provided to the trainer activity must be selected from a pre-defined list of existing datasets in your AI Center instance. You cannot specify a dataset subfolder in the dataset name property of the trainer activity. I think I will proceed with creating a UiPath Support ticket to assist with this. I’m really confused why configuring this is this difficult
Actually, I solved the problem. There was a boolean value set incorrectly which was making the workflow skip the trainer activity. I can produce the training results locally as well as within the AI Center dataset
Great thanks for the update
This topic was automatically closed 3 days after the last reply. New replies are no longer allowed.