Slow run on NF4 using 4060ti (16GB)

[Lordxan]

Finally i can run this model but it takes about 1000s to process one image on 50 timesteps is it somehow possible to optimize?

[neverbiasu]

Model quantization can actually slow down inference and reduce quality.

[Lordxan]

Model quantization can actually slow down inference and reduce quality.

I found one more strange thing. I have about 4 gigs usage when i use NF4 is it normal for 16GB card?

[1blackbar]

ue Dreamo for comfy instead, its similar but faster