Nuke 14.1 and 15 will introduce the possibility to train CopyCat models across multiple machines. I wonder if there are plans to add the support for this in Deadline? As far as I can tell it comes down to running a single command with parameters based on the specific Job on the worker machine, so it shouldn’t be too hard to implement. More information and a sample implementation for OpenCue can be found here: Automated Distributed Training Example
Since CopyCat is getting more and more popular in VFX houses this would be a very nice feature.
Thanks for raising this to us. This looks like a feature request to add in Nuke plugin of Deadline. I will raise an internal ticket to track this request. Unfortunately, this won’t be on the priority list of the feature needs to be added in Deadline.
I’d like to share with you a custom plugin we’ve developed for CopyCat in Nuke. We built it from scratch, and we hope it will be useful to you. You can find the setup guide and the repository at the following link:
The submitter is configured for our specific pipeline scenario, but feel free to customize it to fit your own pipeline needs.
If you need any assistance, don’t hesitate to reach out to me. I’ll do my best to help.