His program uses quantization, but is very optimised and has builds that can fit into 96GB of memory with great results.
DS4 Flash is usually my go-to for a lot of things these days, and I don't have to worry about a cloud model stopping or telling me it's concerned about my usage.
His program uses quantization, but is very optimised and has builds that can fit into 96GB of memory with great results.
DS4 Flash is usually my go-to for a lot of things these days, and I don't have to worry about a cloud model stopping or telling me it's concerned about my usage.