Sebastian Krüger
f27ac50f1d
AudioCraft requires PyTorch 2.1.x. PyTorch 2.4+ changed the torch.utils._pytree API (serialized_type_name parameter), breaking AudioCraft's model loading. - Pin torch>=2.1.0,<2.4.0 - Pin torchaudio>=2.1.0,<2.4.0 - Remove broken pytree monkey-patch attempt After pulling, reinstall with: pip install -r requirements.txt --force-reinstall torch torchaudio 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
31 lines
555 B
Plaintext
31 lines
555 B
Plaintext
# Core ML - Pin to AudioCraft-compatible versions (2.4+ breaks pytree API)
|
|
torch>=2.1.0,<2.4.0
|
|
torchaudio>=2.1.0,<2.4.0
|
|
audiocraft>=1.3.0
|
|
xformers>=0.0.22
|
|
|
|
# UI
|
|
gradio>=4.0.0
|
|
|
|
# API
|
|
fastapi>=0.104.0
|
|
uvicorn[standard]>=0.24.0
|
|
python-multipart>=0.0.6
|
|
|
|
# GPU Monitoring
|
|
nvidia-ml-py>=12.535.0 # Replaces deprecated pynvml
|
|
|
|
# Storage
|
|
aiosqlite>=0.19.0
|
|
|
|
# Configuration
|
|
pydantic>=2.5.0
|
|
pydantic-settings>=2.1.0
|
|
pyyaml>=6.0
|
|
|
|
# Audio Processing
|
|
numpy>=1.24.0,<2.0.0 # NumPy 2.x incompatible with current PyTorch
|
|
scipy>=1.11.0
|
|
librosa>=0.10.0
|
|
soundfile>=0.12.0
|