Quant Request
#1
by
John198
- opened
Hi there, I've been using your quants for a while and they've always been pretty dependable ever since the Midnight Miqu days. Is there a chance you'd be willing to do a 4.5 of https://huggingface.co/Steelskull/L3.3-Damascus-R1? There's big gap between the currently available 4.0 and 5.0 BPW models that could also maximize context while also minimizing perplexity (for those of us using 48VRAM); 5.0 limits to about 32K tokens while 4.5 can still fit 65K.
Dracones
changed discussion status to
closed