fix: honor torchaudio.load(channels_first=False) in patched loader

fix: reject audio shorter than 1.5s with a readable message

debug: surface full traceback in inference error UI

fix: use librosa.load as the single audio loader path

fix: bump torch/torchaudio to 2.6.0 (CVE-2025-32434 mitigation)

fix: pin torch/torchaudio to 2.5.1 (CUDA 12 compatible)

fix: bypass torchaudio backend dispatch for torchaudio 2.9+

fix: guard torchaudio.info patch with hasattr check for torchaudio 2.9+

fix: remove torchcodec, add ffmpeg system pkg, monkey-patch torchaudio to use soundfile backend

fix: correct hardware name to zero-a10g

fix: add torchcodec dep, switch to zero-a10g-small GPU for speed

fix: auto-detect CUDA/CPU device in get_segments_from_wav

fix: use local beat_this checkpoint to avoid download failure in HF Space

fix: pin transformers>=4.46.0 and huggingface_hub to resolve is_offline_mode ImportError

change color for darkmode

change probability func

change probability func

change probability func

change checkpoints

fix some model

fix some model

add hf token

commit for ICASSP 2026 github repository

change description

change description

change desing and result

change import index

delete examples

fix some typo

move audio files to Git LFS

add spaces decorator

initial commit for AIGM

initial segment transformer

initial commit