Problem summary
CopyCat uses VRAM on all GPUs even when limited to not process using those GPUs using --gpu #
Customer reported version
nuke.15.1v4
Customer reported platform
rocky9
Steps to reproduce
1) Setup a basic CopyCat training Nuke script, like the one attached
2) On a multi-gpu machine, open terminal and start monitoring VRAM usage, for example:
watch -n 0.5 -d nvidia-smi
3) Open a terminal and start the Nuke script training using a command like the following:
<nukeInstall> --gpu 0 -X CopyCat1 -x <path/to/nuke/script>.nk
Expected behaviour
For the Nuke to start training the CopyCat node and just use the GPU specified, in this case GPU 0
Actual behaviour
Nuke starts training, and processes on just GPU 0, however it assigns VRAM to all GPUs. So launching a new training session on each GPU would waste more and more VRAM.
Workaround
None
Reproduced by support
This bug has been reproduced in:
Nuke 15.1v4 - Rocky 9.5
Nuke 13.2v8 - Rocky 9.5
*haven't tested os platforms due to setup limitations
Earliest version tested
Nuke 13.2v8 - This issue appears on all tested versions