# of GPUS used to train the model

[seanko29]

Hi!
Thank you for the awesome work.

In the paper, you mentioned that you used RTX 4090 GPUs.
I guess since its plural, you used more than 1?

How many did you use to train the model with 500K iterations and batch size of 32?

Also, does it make a difference when I do the same config but ONLY use 1 GPU? (i guess we see less images when we do that in DDP config?)

[limitlesslab]

@seanko29

you can train the model with a 8 Gb VRAM GPU and get similar results, it's not a heavy architecture and doesn't require vast resources for training and inferencing - I plan to train some models on this architecture myself, feel free to subscribe to my repo and check out the existing models I've already trained