This isn't really anything new; I've been doing something like this for quite a while, I just haven't bothered writing a paper. (: Probably anyone who would ser

This isn't really anything new; I've been doing something like this for quite a while, I just haven't bothered writing a paper. (: Probably anyone who would seriously tackle the problem of "how do I train a huge model on a tiny amount of VRAM?" would come up with something similar. However, most people in the field don't, because the actual practical utility of training huge models on a single GPU is quite low. (e.g they got 341 tok/s for a 14B model on a single 3090 while with my method I was getting ~1k tok/s on a single 4090; that's still very slow) Also, there are more tricks one can use to speed up training/lower VRAM usage which they're not using. For example, you don't need any gradient offloading (you can just accumulate the gradients directly into the optimizers' states if you modify your optimizer), you can use Muon instead of Adam (which needs only half of VRAM of Adam), you can use quantization (both for parameters and for the optimizer states; e.g. I found Muon quantized into 4-bit working relatively well), etc.