Any quantization possible?

#18
by supercharge19 - opened

Can quantized versions be made available, or these models are difficult to quantize?

I don't really know tbh, I think it should probably work with out-of-the-box tools

@supercharge19 Any success with this?

I don't really know tbh, I think it should probably work with out-of-the-box tools

most of deep learning models can be quantized, however, they don't yield good quality outputs, at least not for all outputs (multi lang), that is why onnx models suck.

@anzorq sorry mate, not yet.

Sign up or log in to comment