I know that training deep neural networks (DNNs) takes a lot of computational resources. This is, of course, just a generalized statement. Different networks require different resources.
One that I am particularly interested in is here: https://github.com/NVIDIA/vid2vid
In the research paper, it states that they used:
8 V100 GPUs, each with 16GB memory for training. We distribute the generator computation task to 4 GPUs and the discriminator task to the other 4 GPUs. Training takes ~ 10 days for 2K resolution.
It's understandable why the training process requires a high amount of computational resources. However, once it is trained, can the model run on a robust CPU and a lot of RAM alone?
To be more specific, could it run on AMDs Ryzen 2700x (or Intel equivalent) with 32 GBs of RAM?
I ask because, if this network can run on your typical gaming rig, then there is certainly an application here for a rendering alternative for Video Games. Personally, it would be nice if it could run on a CPU with no help needed from the GPU.
Am I being delusional thinking this can be done with currently available hardware?