Question on Quantized version

[azoksky]

Hi, this is wonderful job to bring it to comfyui. However, with limited resouces I am unable to run 5b version. Since the model has transformer architecture, can't we quantize it to make it fit in low-end systems with 6gb cards? Since gguf formats are omnipreset and users are much more comfortable using it, I guess that is the way forward.