Minimax H3
We need quantized versions of minimax h3. Please make it standard quantization not any of the quality degrading new quantization styles
I don't know what you mean, int8_convrot is the most accurate quantization we've ever used.
Dude tried to hit KJ with a dumb prompt in an unrelated repository, how silly. Int8 and all the other things made have been phenomenal, really should be thankful for comfy, KJ, et al.
Maybe someday they'll catch up and figure out that if gguf were better, they'd be using it.
We need quantized versions of minimax h3. Please make it standard quantization not any of the quality degrading new quantization styles
I suggest you conduct more comparative research on the quantization quality gap between int8 and fp8, then you will understand why people will increasingly prefer int8 quantization going forward.