At the end of the training I have only input models.
How do I change an InputModel into outputModel ?

training_task.connect(opt, name='opt')

yes

"Thanks, it was a cookies issue."

I'm working on an existing code-base, the argparse are named opt instead of args, and the auto connection doesn't work.

I want to connect the argparse manually and load the connected argparse If I'm running on a remote machine.

It working, Thanks!

• I have access to the webUI
• I installed using the docker compose as explained in this tutorial - None

Thanks for the reply.
I wish to save the it in the format:
"output_uri/<pth file>"

I'm trying to use clearml in an existing system and wish to adhere to their conventions

Thanks.
Can we do something similar with artifacts ?

Hi,
I'm using latest clearml-agent version 1.5.2.
This does not always happens, it starts happening after a few runs of successful remote executions.
Then it starts failing, until I remove and rebuild the docker image.

Attaching the clearml.conf

• Thank you, I will give it a try. I debugged the code to investigate the cause of the failure. It appears that the code fails at line 1312 inside Dataset.create during the execution of the instance._serialize() function. I will further explore the code to identify the precise point of failure.