Fix: use wetext on all platforms (avoid pynini dependency on Linux)

Fix: replace WeTextProcessing with wetext (avoid pynini compilation failure)

Fix: add einops, soundfile, vocos, pybase16384, x_transformers

Fix: pin transformers==4.52.1 for OffloadedCache import

Fix: pin torch==2.6.0+cu124, torchaudio==2.6.0+cu124 for CUDA 12 compat

Fix requirements: remove deepspeed, relax torch version for ZeroGPU CUDA 12 compatibility

Upload folder using huggingface_hub

initial commit