v1.3.20 - Train Flux 1 & Flux Klein LoRas

What's new in v1.3.20

1. FLUX.1 dev gets a generate node and LoRA training.
2. FLUX.2 can now train on klein Base 9B as well as 4B, and the Base setting picks which one.
3. 4-bit models now load into RAM and are quantized as they move to the GPU, so peak VRAM drops. FLUX.1 goes from 23.8 to 6.2 GB, and klein 4B from 7.8 to 2.2 GB.
4. App rename from Inline Studio to OpenChar Studio

What's Changed

• Flux lora training: Flux 1 & Flux Klein(4B & 9B) by @imprsnst in #69

Full Changelog: v1.3.19...v1.3.20