Spaces:
Running
on
Zero
Apply for community grant: Academic project (gpu and storage)
Hello, we are releasing the demo for the paper "Text2midi: Generating Symbolic Music from Captions" to be published in AAAI Proceedings.
We would greatly appreciate any support from the Hugging Face community by enabling ZeroGPU and a persistent storage slightly higher than 1GB for uploading our model checkpoint and soundfont files. Thank you.
Hi @keshavbhandari , we've assigned ZeroGPU to this Space. Please check the compatibility and usage sections of this page so your Space can run on ZeroGPU.
As for the persistent storage, I don't think you need it because each Space has 50GB ephemeral storage. https://huggingface.co/docs/hub/en/spaces-storage#persistent-storage-specs
Hi hysts, it seems that zeroGPU is not large enough for the latest model. Are there any alternatives available?
Hi @hysts , what @dorienh meant is - the current GPU time limit of 2 minutes is a bit short for the text2midi model. If it could be extended to 4-5 minutes, users would be able to generate a full minute or minute and half of the audio. Would it be possible to extend the GPU time limit by a few minutes more for generation?
Hi
@keshavbhandari
, thanks for the clarification! But the GPU time limit you mentioned is due to duration=120 in this line. You can set a longer duration, like 180 seconds, if you want. But it's generally not recommended to set a long duration, because ZeroGPU quota is 180 seconds for non-logged-in users and 300 seconds for logged-in users.
It might be better to keep this Space for a quick demo and suggest that users who want to try out the full capability of your method duplicate this Space and allocate paid hardware for their own use.