Flux

This page is a work in progress as information is learned about Flux.

Flux is a DiT Transformer flowmatching model that has high learning potential but it is a very large and also slow model to work with.

Model Details:

Like SD3, a hugginfgace key or a local copy of the diffusers model is needed.

Limitations:

• Finetune is not currently possible.
• Embeddings do not likely work, due to the nature of T5.
• Single file safetensors is currently not supported as a model import
• Full Dora does not currently work completely with the Comfy Lora Loader

Current Information:

• Lora is currently the only recommended training.
• NF4 is possible with the model weights. INT8 requires bits and bytes fixes and is not currently supported.
• The OneTrianer Lora can be used in Comfy in the standard Lora Loader.
• Flux can learn at lower resolutions, for example 512 and 768, which is a benefit for people with less powerful graphics cards.
• It is possible to train a Flux Lora on 12GB. 8GB cards are likely too close to the limit, and would require windows to be using a different GPU for VRAM usage as Windows can quite a bit of VRAM.