Spaces:
Running on Zero
Running on Zero
Commit History
-
[Admin maintenance] Support new ZeroGPU hardware Running
Verified signaturea6d4a33
-
Add Advanced tab mirroring reference repo UI
Wraps the existing UI in a Simple tab and adds an Advanced tab that mirrors stable_audio_3/interface/diffusion_cond.py: negative prompt, sampler params (sigma_max, APG, duration padding), init audio + noise level, inpainting (audio + mask start/end), output spectrogram gallery, and send-to-init / send-to-inpaint buttons. SAMPLERS narrowed to those valid for rf_denoiser. Inpaint/init audio is pre-resampled to model SR and cast to model dtype to avoid fp16/fp32 mismatches in the pretransform encoder. matplotlib/Pillow added for the mel-spectrogram helper.232ab2a
-
Cosmetic tweaks
d276c0e -
Fix Blackwell support: pin numpy<2, install stable-audio-tools --no-deps, fall back to SDPA
5728513 -
Update app.py
Verified signatureb134d23 -
Update requirements.txt
Verified signature74a1656 -
Update app.py
Verified signaturebad2e14 -
Update requirements.txt
Verified signature51a5748 -
Update requirements.txt
Verified signature5d9a11e -
Update requirements.txt
Verified signature8828527 -
Update requirements.txt
Verified signature31189fb -
Update requirements.txt
Verified signaturecfb6267 -
swap torchaudio -> soundfile (avoid libcudart.so.12 load failure)
Verified signatured82866b -
install flash-attn + stable-audio-tools at startup (SKIP_CUDA_BUILD)
Verified signature72c410e -
pin PyWavelets>=1.7 for NumPy 2.x ABI
Verified signaturecba3206 -
initial commit: SA3 medium + small-music + small-sfx
Verified signature5e77923 -
initial commit
Verified signature366a3bb