stable-diffusion
stable-diffusion copied to clipboard
Upgrade to Lightning 1.7
Hey! Needless to say incredible work with Stable Diffusion and latent diffusion in general.
I saw Stable Diffusion is using a old-ish version of PyTorch Lightning (1.4.2), I'm wondering if you'd like help upgrading to Lightning 1.7, happy to provide it. The idea would be to create a test, ensure there's (at least) parity on results and upgrade.
Here's a breakdown of what was released since 1.4.x just in case:
- 1.5 LightningLite, Fault-Tolerant Training, Loop Customization, Lightning Tutorials, LightningCLI v2, RichProgressBar, CheckpointIO Plugin, and Trainer Strategy Flag
- 1.6 Support Intel's Habana Accelerator, New efficient DDP strategy (Bagua), Manual Fault-tolerance, Stability and Reliability
- 1.7 Apple Silicon support, Native FSDP, Collaborative training, and multi-GPU support with Jupyter notebooks
Extra note: Lighting 1.7 supports PyTorch 1.9+