When running on a Mac with MPS available, by modifying the remwm.py file, the code correctly detects MPS as available and loads Florence-2 model to MPS and the LaMa model processing falls back to CPU, which creates a bottleneck, where the second pass would take significant time. May I know is it possible that MPS is available to be used in the second pass?
Also, I would like to know whether it can detect multiple watermarks. Thank you so much
When running on a Mac with MPS available, by modifying the remwm.py file, the code correctly detects MPS as available and loads Florence-2 model to MPS and the LaMa model processing falls back to CPU, which creates a bottleneck, where the second pass would take significant time. May I know is it possible that MPS is available to be used in the second pass?
Also, I would like to know whether it can detect multiple watermarks. Thank you so much