Spaces:
Running
on
L4
Apply for community grant: Academic project (gpu)
Hello,
We're excited to share the HF Demo of our HSMR for Human Biomechanical Skeleton and Mesh Reconstruction from monocular images has been developed. To maintain efficient public reproduction to this demonstration, we're applying for GPU resource support.
This follows our successful deployment of prior research (HMR 2.0, HaMeR and GVHMR) on Hugging Face platforms, made possible through a community grant's generous support. We hope HSMR can get similar supports.
We'd be happy to provide additional information as needed. Thank you for considering our resource request.
Best,
Yan Xia
Hi
@IsshikiHugh
I couldn't find your grant request in our internal Slack channel, so I believe it was submitted without using the button in the Space Settings (see docs). This button sends a notification to our internal Slack channel, where grant requests are reviewed by the HF staff.
Please note that even HF staff members who have permission to access private Spaces do not receive notifications from private Spaces. So, if the button is not used, we may not be aware of your request.
We would appreciate it if you could use the button next time when submitting a request.
Anyway, I just assigned L4 for now, but would it be possible to migrate this Space to use ZeroGPU?
Hi
@hysts
Thank you very much! I just follow what I can see in previous repositories so I don't know there is a button. But I will do that next time!
Regarding the GPU, can the migration be a little bit later? I used to use ZeroGPU for developing GVHMR but it's hard to iterate the version due to the limited GPU quant. (Or do you have any suggestions for developing with ZeroGPU? Sometimes I need to adjust the allocation of the quant and it wastes my quant.)
I will notify you when I get some time to implement that and finish the adaption. Thank you very much!
Regarding the GPU, can the migration be a little bit later?
Sure, no problem. There's no need to migrate right away. It would be great in the long run if it could be migrated to ZeroGPU, but if it's difficult, it's just fine.
I used to use ZeroGPU for developing GVHMR but it's hard to iterate the version due to the limited GPU quant. (Or do you have any suggestions for developing with ZeroGPU? Sometimes I need to adjust the allocation of the quant and it wastes my quant.)
Yeah, I understand the frustration, but unfortunately, there's no good solution for the ZeroGPU quota limit. (Even HF staff are affected by the same limit.)
Get it! Thank you!