Switch to CUDA base image + cu121 torch to use GPU

Fix AV1/unsupported codec: transcode input via ffmpeg before OpenCV decoding; fix OMP_NUM_THREADS warning

Fix torchvision 0.16+ compat: use functional instead of functional_tensor

Switch to Docker SDK: fix build order so basicsr_vendor exists before pip install

Vendor patched basicsr to fix Python 3.13 get_version() build error

Add Real-ESRGAN upscaler app

initial commit