lucascs@protonmail.com Lucas C.S. Audio transforms, filters, forced alignment and RNN-T loss for PyTorch. Upstream entered a maintenance phase with 2.9 and removed its own media I/O; loading and saving now go through sci-ml/torchcodec. Build the CUDA kernels (RNN-T loss, forced alignment, IIR filter) and the CUDA CTC decoder pytorch/audio torchaudio https://github.com/pytorch/audio/issues