Release 1.6 Starlight model sizes

I didn’t check the sizes in rel. 1.5, but compared to 1.4 the neuroserver models have all gotten about 1% larger:

Model 1.4 Size 1.6 Size
SL Mini 2.8GB 3GB
SL Sharp 2.9GB 3.1GB
SL Fast 6GB 6.4GB
SL High Quality 2.8GB 3GB
SL Precise 2.5 6.8GB 7.3GB

So it is probably worth retesting them all on your system.

We’ll definitely do.

But first it’s testing the new Mac version :+1:

No change in functionality for me:

SL Mini, HQ and Precise run correctly, Sharp go off into a loop before creating an output file and Fast throws an inference error after loading the model.