mradermacher's picture
auto-patch README.md
b41562a verified
|
raw
history blame
3.12 kB
metadata
base_model:
  - Himitsui/Kaiju-11B
  - Sao10K/Fimbulvetr-11B-v2
  - decapoda-research/Antares-11b-v2
  - beberik/Nyxene-v3-11B
language:
  - en
library_name: transformers
license: apache-2.0
quantized_by: mradermacher
tags:
  - moe
  - frankenmoe
  - merge
  - mergekit
  - Himitsui/Kaiju-11B
  - Sao10K/Fimbulvetr-11B-v2
  - decapoda-research/Antares-11b-v2
  - beberik/Nyxene-v3-11B

About

weighted/imatrix quants of https://huggingface.co/Steelskull/Umbra-v3-MoE-4x11b

static quants are available at https://huggingface.co/mradermacher/Umbra-v3-MoE-4x11b-GGUF

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 7.7 for the desperate
GGUF i1-IQ2_XXS 9.8
GGUF i1-IQ2_XS 10.9
GGUF i1-IQ2_S 11.1
GGUF i1-IQ2_M 12.2
GGUF i1-Q2_K 13.4 IQ3_XXS probably better
GGUF i1-IQ3_XXS 14.2 fast, lower quality
GGUF i1-IQ3_XS 14.9
GGUF i1-IQ3_S 15.8 fast, beats Q3_K*
GGUF i1-IQ3_M 16.1
GGUF i1-Q3_K_M 17.5 IQ3_S probably better
GGUF i1-Q4_K_M 22.1 fast, medium quality

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