Open Dan-Kouba opened 1 year ago
fixed in latest commit, but you're passing a jpg image as a checkpoint. You need to pass a checkpoint path containing the pretrained weights for the models.
Also, haven't tested, but this is unlikely to work on osx. You can try by setting device to "mps", but will likely fail in silent/unexpected ways. You can also try running on cpu device, which will be v slow.
Thanks for the insight! This was my first time trying to run the command and I wasn't clear on how to use all the arguments. Is the checkpoint something that is included or is it something I need to generate myself?
I have the same question about the checkpoint, because in README.md there is nothing about it
I can try to find the checkpoint, but this was just a side project, had no idea it would get this popular haha 😅 Otherwise, you can train your own checkpoint with the training script. Remember, this isn't production code, it's just a research project so there may a few issues you have to debug
No problem at all! I'm happy to play around now that I know more about the requirements to run the script. Thanks.
I installed everything on OSX, but I get this error when I try to run the program:
dankouba@dan-koubas-macbook-pro:~/Documents/personal/Color-diffusion$ python3 inference.py --image-path ~/Downloads/md.jpg --checkpoint ./checkpoint/md.jpg --save_path ./output Traceback (most recent call last): File "/Users/dankouba/Documents/personal/Color-diffusion/inference.py", line 37, in
encoder = Encoder(**enc_config)
^^^^^^^^^^^^^^^^^^^^^
TypeError: Encoder.init() got an unexpected keyword argument 'condition'
My guess is one of the requirements needs to be pinned to a version, but they all take up the latest. Any insight you can provide?