Custom Models

[Cyboghostginx]

Can't one just convert model like below?

from mlx_lm import convert
convert("black-forest-labs/FLUX.1-dev", quantize="4bit", output_dir="flux_q4_mlx")

then use that model manually with your node instead of using the 3 fixed models?

[PlebeiusGaragicus]

I had this same question today… will be looking into it. DiffusionKit page describes converting/quantizing models.