●Change8

v5.7.2rc1

📦 invokeaiView on GitHub →
✨ 2 features🐛 2 fixes🔧 1 symbols

Summary

This release introduces an optional configuration to switch to the CUDA memory allocator for better VRAM management and performance, alongside support for WEBP image uploads and several UI fixes.

Migration Steps

  1. If you wish to use the CUDA memory allocator for potential VRAM improvements, add the `pytorch_cuda_alloc_conf` setting to your `invokeai.yaml` file (e.g., `pytorch_cuda_alloc_conf: "backend:cudaMallocAsync"`).
  2. The Invoke Launcher is now the recommended installation method. Follow the Quick Start guide to use it, or use the manual install guide for headless setups.

✨ New Features

  • Added support for uploading WEBP images, which are converted to PNGs internally.
  • Introduced the `pytorch_cuda_alloc_conf` setting in `invokeai.yaml` to allow opting into CUDA's memory allocator for potentially reduced peak VRAM usage and improved performance.

🐛 Bug Fixes

  • Fixed rendering issue in the Workflow Editor for "single or collection" field types, resolving display problems for widgets like IP Adapter images and ControlNet control weights.
  • Corrected the download button in the Workflow Library list to download the intended workflow instead of the currently active one.

🔧 Affected Symbols

pytorch_cuda_alloc_conf