check this one out

[adrianpuiu]

https://research.nvidia.com/labs/par/Perfusion/ , code coming out soon from nvidia

[seruva19]

Wow, that's cool stuff. Even though it is claimed to require 27 Gb of VRAM for training 😑

[adrianpuiu]

for training i'm covered. i got 48Gigs of Vram .. but having 100 kb models , hah , that will save my hard drives space big time :))

+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 530.41.03 Driver Version: 530.41.03 CUDA Version: 12.1 |
|-----------------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+======================+======================|
| 0 NVIDIA GeForce RTX 4090 Off| 00000000:01:00.0 Off | N/A |
| 0% 40C P8 41W / 450W| 8MiB / 24576MiB | 0% Default |
| | | N/A |
+-----------------------------------------+----------------------+----------------------+
| 1 NVIDIA GeForce RTX 4090 Off| 00000000:02:00.0 On | Off |
| 0% 42C P5 35W / 450W| 8808MiB / 24564MiB | 22% Default |
| | | N/A |
+-----------------------------------------+----------------------+----------------------+