Hi PerfectMole86 ,

How are you saving your models? How are they being saved, under .pipeline folder as well?

I have used a output_uri argument in my Task initialization for storing my models

My models are not getting saved in the .pipeline folder. They are not getting saved in the output_uri specified in the Task.init as well.

The model is only getting saved locally

I got this message when the training started and it is only saving the model locally

How is the model being saved/logged into clearml?

It was getting logged when I was running it as an individual task but it is not getting logged in the pipeline. I was plotting loss graphs too but now they're not getting plotted.

Do I have to run a task individually completely for it to function properly in a pipeline?

What if you point it to the fileserver? Does it still not upload the model?

I haven't pointed it to the file server because I'm running a locally deployed docker instance of ClearML

This is how I am defining the task in the code

and I'm plotting the losses like this

And how is the model being saved?

Just locally but it is saved fine

I mean code wise. Also where is it saved locally?

Not on the NAS storage but on my PC where the training is running

How is the model saved in the code?

Is output_uri defined for both steps? Just making sure.