mradermacher's picture
auto-patch README.md
8168320 verified
|
raw
history blame
4.33 kB
metadata
exported_from: cloudyu/Mixtral_11Bx2_MoE_19B
language:
  - en
library_name: transformers
license: cc-by-nc-4.0
quantized_by: mradermacher

About

weighted/imatrix quants of https://huggingface.co/cloudyu/Mixtral_11Bx2_MoE_19B

Usage

If you are unsure how to use GGUF files, refer to one of TheBloke's READMEs for more details, including on how to concatenate multi-part files.

Provided Quants

(sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants)

Link Type Size/GB Notes
GGUF i1-IQ1_S 4.3 for the desperate
GGUF i1-IQ2_XXS 5.4
GGUF i1-IQ2_XS 6.0
GGUF i1-IQ2_S 6.2
GGUF i1-IQ2_M 6.7
GGUF i1-Q2_K 7.3 IQ3_XXS probably better
GGUF i1-IQ3_XXS 7.7 lower quality
GGUF i1-IQ3_XS 8.1
GGUF i1-Q3_K_S 8.6 IQ3_XS probably better
GGUF i1-IQ3_S 8.6 beats Q3_K*
GGUF i1-IQ3_M 8.8
GGUF i1-Q3_K_M 9.5 IQ3_S probably better
GGUF i1-Q3_K_L 10.3 IQ3_M probably better
GGUF i1-IQ4_XS 10.5
GGUF i1-Q4_0 11.1
GGUF i1-Q4_K_S 11.2 optimal size/speed/quality
GGUF i1-Q4_K_M 11.8 fast, recommended
GGUF i1-Q5_K_S 13.5
GGUF i1-Q5_K_M 13.9
GGUF i1-Q6_K 16.0 practically like static Q6_K

Here is a handy graph by ikawrakow comparing some lower-quality quant types (lower is better):

image.png

And here are Artefact2's thoughts on the matter: https://gist.github.com/Artefact2/b5f810600771265fc1e39442288e8ec9

Thanks

I thank my company, nethype GmbH, for letting me use its servers and providing upgrades to my workstation to enable this work in my free time.