* feat: Enhance BAGEL Model Loader with Dynamic Quantization Support
- Added support for dynamic quantization modes (BF16, NF4, INT8) in BagelModelLoader.
- Updated INPUT_TYPES to include quantization options and advanced settings.
- Enhanced VALIDATE_INPUTS to validate quantization modes and check for bitsandbytes availability.
- Modified load_model to handle different quantization modes and improved memory management.
- Updated model loading logic to accommodate quantization configurations.
- Adjusted cfg_renorm_min default values in BagelTextToImage and BagelImageEdit classes.
- Added bitsandbytes as a requirement in requirements.txt.
* feat: Add optimal memory calculation for GPU based on quantization mode
* feat: Update README and pyproject.toml for advanced quantization support and version bump to 1.0.4
* fix: Revise device_map
* refactor: Clean up comments and improve code readability in BagelModelLoader
* feat: Update BAGEL workflows with enhanced node configurations and improved properties
* feat: Add offload directory creation for model dispatch in BagelModelLoader
绯世 Haise
e806fc1ca0
