v0.3.46
📦 comfyuiView on GitHub →
✨ 10 features🐛 5 fixes🔧 3 symbols
Summary
This release introduces significant hardware support updates, including Iluvatar CoreX and WAN 2.2, alongside improvements to XPU functionality and training node features. It also replaces torchaudio.load with pyav for audio loading.
✨ New Features
- Added support for Iluvatar CoreX.
- Added WAN ATI support.
- Added WAN 2.2 support.
- Added Moonvalley Marey V2V node with updated input validation.
- Added map_function to get_history.
- Added default device argument.
- Enabled pytorch attention by default for gfx1201 on torch 2.8.
- Added xpu version for async offload and some other things.
- Separated denoised and noise estimation in Euler CFG++.
- Training Node now supports algorithm selection, gradient accumulation, and optional gradient checkpointing.
🐛 Bug Fixes
- Fixed issue where xpu function was not implemented (multiple fixes).
- Fixed issue with line endings in the GitHub workflow.
- Only enable cuda malloc when using cuda torch.
- Updated negative prompt for Moonvalley nodes.
- Fixed link from docs due to document updates.
🔧 Affected Symbols
torchaudio.load/system_stats APIEuler CFG++