Hugging Face's logo

mo137
/
Amethyst-13B-Mistral-8bpw-hb8-exl2

Text Generation
Transformers
Safetensors
llama
exllamav2
exl2
Text Generation
Not-For-All-Audiences
nsfw
Transformers
text-generation-inference
Inference Endpoints
Model card Files Community
Amethyst-13B-Mistral-8bpw-hb8-exl2 / README.md
mo137's picture
mo137
Update README.md
761df1a
preview code
|
raw
history blame
1.01 kB
metadata
license: cc-by-nc-4.0
tags:
  - exllamav2
  - exl2
  - Text Generation
  - not-for-all-audiences
  - nsfw
  - Transformers
  - llama
  - text-generation-inference

Amethyst 13B Mistral - EXL2 - 8bpw, hb8

Description

  • 8 bits per weight.
  • 8 bits "for the lm_head (output) layer of the model," instead of the typical 6.
  • Works fine with 24 GB VRAM and no flash attention v2 under Windows.
  • For me runs at about 64% of the 4-bit GPTQ speed.

I converted the model using the convert.py script from the exllamav2 repo:
https://github.com/turboderp/exllamav2
Its documentation:
https://github.com/turboderp/exllamav2/blob/master/doc/convert.md

Measuring the model took 51 minutes, converting it 18 minutes.

I used the WikiText-2-v1 dataset for calibration:
https://huggingface.co/datasets/wikitext/blob/refs%2Fconvert%2Fparquet/wikitext-2-v1/test/0000.parquet