Spaces:
Running
on
Zero
Apply for community grant: Academic project (gpu)
We would like to release a public demo of the upcoming COP-GEN-Beta model: https://miquel-espinosa.github.io/cop-gen-beta/
It is the first of a kind, flexible diffusion model trained densely on the Major TOM thumbnails from across the globe. It is capable of reconstrucing any subset of supported modalities (optical, radar, elevation) and can also perform unconditional generation.
If possible, we would like to request a T4 access to allow the users to run multiple generations to test out the model.
Hi @mikonvergence , does ZeroGPU not work for this Space? I've assigned ZeroGPU to this Space for now.
Hello @hysts - ZeroGPU should work functionally (I still need to integrate the backend before tomorrow in fact), but I hoped we could get an instance that allows the users to explore the model under less severe compute usage restrictions.
We expect this model to be quite popular and wanted the user community to be able to explore different synthesis configurations of the model in the app, which likely exceeds the current usage limit of ZeroGPU.
Thanks for the clarification! I understand that some users might want to go beyond their ZeroGPU quota, but they can duplicate the Space and assign paid hardware, or subscribe to PRO to increase their quota. Also, they can just download the Docker image of this Space and run it on their own environment. The main goal of the Space grant is to let as many people as possible quickly try out demos, and it is not meant to give unlimited free inference time to heavy users. Thanks for your understanding.
Understood, thanks!