Some additional quants
Just FYI, I've added some additional static and imatrix quants at https://huggingface.co/mradermacher/QuartetAnemoi-70B-t0.0001-i1-GGUF and https://huggingface.co/mradermacher/QuartetAnemoi-70B-t0.0001-GGUF
They should provide the full set of quantization options, including IQ1_S and IQ4_NL, for what it's worth. They are not meant to replace existing quants, just to add more options.
And if I may say so, this is an impressive model so far. It follows my instructions so precisely that I frequently have to explicitly ask it to embellish stuff or add more on its own. Extremely good for developing stories (and probably lots of other things).
Thanksl I'll add them to the card. I tried the 1.73 bpw quant out of curiosity and while it's not good, it's better than totally broken output which was the expectation I had for something under 2 bits.
Yeah, the IQ1_S quants are... really not something I'd like to use myself, but they work surprisingly well, especially for larger models (120bp+) and shorter contexts. I provide them to allow the most people access to models. And my weight matrices are probably also not optimal (certainly not for non-english uses), but quantization algorithms only improve over time...