[00:00:01] Hi everyone. In this video, we'll talk about tasks that are performed when a job is actually executed.
[00:00:09] Here, I have a job that has been completed. If we go here into Actions, View tasks, you see the number of steps that were performed in order to synthesize the data set.
[00:00:20] Let's have a look at those. The first one is train data model. The data is fetched, it's analyzed, it's encoded because it needs to be brought into the proper format to then go into the training of our AI model.
[00:00:33] Here actually, if we look into details, we can see the details of how the model was trained. We see the validation loss, so how well the model learned, the patterns of the existing data, and that is decreasing which is good. At this point, it starts to increase which is not so good.
[00:00:52] Basically, this minimum validation loss was selected, because it has the lowest validation loss.
[00:01:02] Once the training has been completed, data is generated for the model QA. We have two generation steps actually. One is data that is here created for the Model QA report, and that's also analyzed for the Model QA report.
[00:01:18] Then we have the second step here, generating data for delivery and for the Data QA report.
[00:01:25] Then, there's some post-processing steps, exporting of the data, packaging, and everything is packaged into a zip archive, the data itself but also the QA reports, and then everything is completed.
[00:01:38] This was a very fast job. It took here, basically, just a couple of seconds to complete. For larger data sets, this can obviously take much, much longer.
[00:01:49] Thanks for watching.