I'll be pausing the space due to high costs in keeping it live

#3
by rdesai2 - opened

Please DM on Twitter/X if you'd like me to turn it back on.

can this be used without comfy ?

No, currently it cannot. See this GitHub issue for a detailed explanation.

Hi @rdesai2 ! sorry to hear this, but I get it tho :(

I tried to duplicate it following the recomendation from Hugging but fail to run it due to Torch not being found. I'm not an expert with Hugging, but could be due to the fact that im using the Basic CPU machine? Or there is acttually something missing in the code?

Traceback (most recent call last):
File "/home/user/app/demo.py", line 17, in
from test import process_face
File "/home/user/app/test.py", line 4, in
from face_enhance import face_enhance
File "/home/user/app/face_enhance.py", line 5, in
import torch
ModuleNotFoundError: No module named 'torch'

There are a few issues here

  1. This requires a GPU (e.g. 1 L40S or HF's ZeroGPU)
  2. This requires persistent storage (HF's medium storage) due to the large model weight sizes and
  3. HF_DEMO=true needs to be set in your HF space's environmental variables. Running this elsewhere you can ignore this.

I think rather than duplicating the HF space, you should rent a GPU instance from RunPod and follow the instructions in the Github README file. It's not free, but getting started is very quick for just a few dollars.

@DomingoSL I've temporarily turned the space back on - let me know what you think!

@rdesai2 yeah I was thinking about runpod too, but the problem is the huge space the model needs, Runpod have reasonable prices for the GPU consumption but for the disk not so much :(
Thanks again for making it available again tho!

Your need to confirm your account before you can post a new comment.

Sign up or log in to comment