Smaller model training scripts

[josht000]

Do you have suggestions for using smaller models and/or quantized models for SVF? Any results or experirences you can share during your research on these? We're having a hard time getting 2xH200 GPUs for any length of time.

Thanks!

[BradKML]

Seconding this, but further towards consumer GPUs