How to use Upload File activity?

Help AI Center
, , , ,
1

Hi, I need some help about an activity “Upload File(from MLService Activities)”
Is it possible to improve the ML skill using this activity?
Also, can you use it with ML Extractor Trainer at the same time or use it as replacement of ML Extractor Trainer?
If so, please tell me how to use this activity. I NEED TO KNOW what blanks on properties I should fill in.

image
image1661×435 82.7 KB

2

Hi @rtsutsui, welcome to the Community.

Did you go through the documentation of this activity? It will help you with the properties of the activity & usage:

https://docs.uipath.com/activities/other/latest/user-guide/ml-storage-upload

Best Regards.

3

@rtsutsui

This is used to inprove the dataset or add new data to the dataset to be trained…

for example if you have a text classification model which is working with 90 percent accuracy…you can enable auto retrain in your ai center and when new data comes in after the manual validation you can collect the data and add the new file to the ai center dataset so that next time the ml skill auto trains then the new file also would be included in the training as the data is uploaded to the training dataset

The path is relative path in the dataset created in ai center…and file is the local file to be uploaded

cheers

4

@arjunshenoy Thank you for your response. Yes I did check the documentation but unfortunately, I could not find any examples to use this activity like what file I am going to upload through “Upload File” and how it is going to affect the ML skill.
I hope you do not mind sharing a bit more detail about this activity since I could not figure out how to use and find anyone actually using this activity.
I have been trying to fix it to work and I will share the properties and hope it will help.
image

5

@Anil_G Thank you for your response. Can you please tell me what file I am supposed to upload through this activity?

image
image1021×397 12.1 KB

And I assume in this case, the relative path is “Datasets/receipt” and going to be set as same class as “export” and “fine-tune” but as you may know, you can not choose folders if it includes “fine-tune” as input dataset to run pipelines so should it make inside export like “Datasets/receipt/export”?

6

@rtsutsui

You would upload the documents to the required dataset …for example…the dataset for placing the documents which are being trained is the train folder then yes the new files also would be uploaded to the same…

Again…this option best suites for models where datalabeling is not required…as stated above example is the text classification model…where the training input required is two columns in a csv file one for data and one for outcome…

For the du models where labeling is required it is preferred to use ml learning training extractor which would auto upload the dataset and the export that has been validated by the human in case of failed or manual cases…if you just upload the document the retraining model would not get teained as the labeling or the expected values outcome is missing

Cheers