Blake

Flux LORA weight effect differs when using Q8_0 GGUF and FP8 safetensor

3

LORAs do not seem to apply the same weights when using flux1-dev-fp8.sft as they do when using flux1-dev-Q8_0.gguf. Investigated this a bit and it seems to be happening in every...

LORAs not working with flux1-dev-Q8-0.gguf

7

Since commit 13d6f8ed900b0857e9872e67befb02f7ed54da35 LORAs don't seem to work with the flux1-dev-Q8-0.gguf checkpoint. Generating without a LORA seems fine, but adding one causes the error in the attached txt file. Tried...

GGUF Q8_0 and Automatic Diffusion (8bit LoRa) generating poor quality

4

Since ba01ad37, LoRas loaded in 8bit to the Q8_0 GGUF generate to a poor quality. Loading the LoRa in 16bit appears to fix this issue, but there are subtle differences...

Tested on windows?

7

Was keen to give this a try as it looks very cool, but ran into issues on windows. Install gives hardlink error ``` F:\Software\AI\GlaDOS>python scripts\install.py UV is already installed Using...