1.2.0

• Update backend to Python 3.12 and PyTorch 2.5 (NOTE: updating the backend may take several minutes when launching this version of the program for the first time)
• Performance optimizations:
  • New Version (1.2.0): 42 seconds to 4x upscale 35 1400p grayscale images saved with PNG format
  • Old Version (1.1.8): 64 seconds
  • Use pyvips to speed up image encoding and decoding
  • Use channels last format to speed up model inference
  • Use bf16 (if supported by GPU) when upscaling with DAT model