This repository has been archived by the owner on Dec 14, 2023. It is now read-only.
Tags: ExponentialML/Text-To-Video-Finetuning
Tags
Merge pull request #26 from ExponentialML/version/version2 Text To Video Finetuning Version 2 ## Changes and Updates - [x] High quality VRAM config. - [x] Add text encoder training. - [x] Allow training on low vram systems. - [x] Allow single image training. - [x] Train with image captions. - [x] Train with video captions in folder. - [x] Gradient checkpointing support. - [x] Time agnostic training. - [x] Add aspect ratio bucketing. - [x] Verify installation. - [x] Add hybrid LoRA for training. - [x] Add latent caching. - [x] Add optimizer agnostic settings in config. - [x] Soup up unet finetuner for readability and efficiency. - [x] Update README to reflect training.