Eric Rollei
c36f3e6d7c
New nodes: - Hunyuan Instruct Loader/Generate/ImageEdit/MultiFusion/Unload - Block Swap for GPU<->CPU transformer block swapping - HighRes Efficient with loop-based MoE routing - Unified V2 auto-detecting generate node - VAE Simple, Memory Budget, Memory Manager, Cache V2 Changes: - Centralized model path system via folder_paths (hunyuan + hunyuan_instruct) - INT8 bitsandbytes fix for CB/SCB preservation during .to() calls - All loaders migrated to resolve_hunyuan_model_path() - Updated README with Instruct docs, HF links, hardware tables - pyproject.toml bumped to v1.1.0, deps synced - Moved dev docs to Docs/, cleaned .gitignore Quantization: - Added Instruct INT8/NF4 quantization scripts (distil + full) - Added HF upload script for Instruct models
Example Gallery
Use this directory to store curated renders that demonstrate different loader/generator combos and prompt rewriting modes.
Recommended naming: workflow-name_resolution_seed.png with a matching .md snippet describing prompt, steps, sampler, and any notable settings.