Fix: replace torch==2.1.0 with torch>=2.5.0 (no 2.1.0 wheels available)

Update demos/musicgen_app.py

Update demos/musicgen_app.py

Update demos/musicgen_app.py

Update demos/musicgen_app.py

Update demos/musicgen_app.py

Update demos/musicgen_app.py

syntax and indentation errors correction

Fix indentation error at line 414 in ui_batched function

Increase @spaces.GPU duration from 120s to 300s

Fix batched UI IndentationError: add 4-space indent to lines 413-414

Fix IndentationError: add 4-space indent to output components (lines 283-286) inside with gr.Column() block

Fix TabError: replace tab chars with spaces on line 283 output component

Fix IndentationError: use tab indentation on output components to match surrounding code style

Fix AttributeError: replace gr.make_waveform with direct wav passthrough; replace gr.Video outputs with gr.Audio

Fix AssertionError: replace ProcessPoolExecutor with ThreadPoolExecutor for ZeroGPU compatibility

Increase @spaces.GPU duration to 120s and set default model to musicgen-small

Add @spaces.GPU to predict_full to fix CUDA init error

Fix: pin transformers<=4.45.2 to fix register_pytree_node AttributeError with torch 2.1.0

Fix: accept torch==2.1.0 (audiocraft requirement), transformers<5.0, encodec+audiocraft from git

Fix: encodec from git (broken setup.py in PyPI), audiocraft from git, transformers<5.0.0

Fix: pin torch==2.9.1 explicitly first to prevent audiocraft from downgrading it

Fix: switch to audiocraft from PyPI (no hard torch pin), remove spacy/demucs direct pins to prevent torch downgrade

Fix: remove xformers entirely (no release supports torch 2.9.1, was downgrading torch)

Fix: remove torchaudio/vision/text pins (HF pre-installs torch 2.9.1), fix xformers + spacy yanked

Fix: add import spaces + @spaces.GPU decorator on _do_predictions for ZeroGPU detection

Fix: pin transformers<5.0.0 (v5 requires torch>=2.4, ZeroGPU has torch 2.1.0)

Fix: proper requirements.txt for Gradio SDK - no torch pin (HF pre-installs it), full audiocraft deps with av==11.0.0

Fix: add FFmpeg dev libs and pkg-config to packages.txt for av==11.0.0 wheel build

Fix: revert to sdk: gradio (required for ZeroGPU), restore app_file

Fix: remove app_file (invalid for Docker SDK) to clear configuration error

Fix: switch sdk from gradio to docker so custom Dockerfile is used

Fix: restore full audiocraft-compatible deps with exact pins (av==11.0.0, xformers<0.0.23, demucs, all missing packages)

Fix: remove demucs and xformers (source of av build failure)

Fix: add FFmpeg dev libs (libavformat-dev etc.) to fix av wheel build failure

Fix Dockerfile: use slim base, add git/ffmpeg, PyTorch extra-index-url, non-root user

Fix requirements: pin torch, add audiocraft from GitHub, add all deps

Upload requirements.txt with huggingface_hub

Upload requirements.txt with huggingface_hub

Upload requirements.txt with huggingface_hub

Upload requirements.txt with huggingface_hub

Upload requirements.txt with huggingface_hub

Upload requirements.txt with huggingface_hub

Upload packages.txt with huggingface_hub

Upload README.md with huggingface_hub

Upload requirements.txt with huggingface_hub

Upload folder using huggingface_hub

Upload app.py with huggingface_hub

Upload README.md with huggingface_hub

Deploy: Simple Flask App Test