Spaces:
Running
on
Zero
🚩 Report
IU want to you do increase gpu quota.
The ZeroGPU team added a dynamic quota configuration a couple of weeks ago, so I incorporated it now. It should work better now.
I sadly can't increase the total GPU quota, since it's determined by the Hugging Face team. But now if you do reach the quota, you can lower the number of steps and it will dynamically require less resources.
You can still duplicate the space and run it without limits, just slower, the ZeroGPU interface should not break things if you don't have a paid GPU plan (of course you can also pay for an added GPU on your own space).
Hey
@deeplearning
, I added a parallel batch option so that it will take a bit less time, especially for heavier models& lots of timesteps.
You can also run this space locally on your own computer if you own a CUDA compatible GPU.
The ZeroGPU quota is still not in my control. Please close this issue if everything else is ok :)