Apply for community grant: Academic project (gpu)
Hi, I'm Xueqing Wu, a first-year PhD student at UCLA, and also the first author of this paper: https://arxiv.org/abs/2406.13444
The VDebugger demo is for this research paper, and we are planning to make it public very soon after some final adjustments. We will be extremely grateful if we could get some free GPU!
Additionally, we would also really love it if we could possibly get an upgrade of GPU. This project requires loading multiple models into the GPU, so the RAM limitation is extremely hard for us. We have to make a few compromises such as using 4-bit quantization to make it possible. If we could get a GPU with larger RAM or get multiple GPUs then that would be extremely helpful.
Thank you in advance!
Thank you for the quick response! We considered using ZeroGPU at the beginning, but since our organization is not Enterprise organization, I don't think we can access ZeroGPU.
About the space building speed, my guess is the space needs to load multiple models which takes a long time. After loading the model, it takes ~4min to run each example.
Thanks for the reply. I see.
As we use ZeroGPU as the default hardware for grants, it would be nice if you could migrate your Space to ZeroGPU if it works on it. But if it takes about 4 minutes to run, maybe it's not a good idea to use Zero anyway.
I'll assign 1xL4 to this Space for now, then.
Thank you so much! We'll keep trying optimizing the demo. We may reach out again when we figure out how to make it faster or migrate to ZeroGPU.