bartowski/Cerebrum-1.0-8x7b-exl2

Exllama v2 Quantizations of Cerebrum-1.0-8x7b

Using turboderp's ExLlamaV2 v0.0.16 for quantization.

The "main" branch only contains the measurement.json, download one of the other branches for the model (see below)

Each branch contains an individual bits per weight, with the main one containing only the meaurement.json for further conversions.

Original model: https://huggingface.co/AetherResearch/Cerebrum-1.0-8x7b

Branch	Bits	lm_head bits	VRAM (4k)	VRAM (16k)	VRAM (32k)	Description
6_5	6.5	8.0	38.9 GB	40.4 GB	42.4 GB	Near unquantized performance at vastly reduced size, recommended.
4_25	4.25	6.0	25.9 GB	27.4 GB	29.4 GB	GPTQ equivalent bits per weight, slightly higher quality.
3_75	3.5	6.0	23.0 GB	24.5 GB	26.5 GB	Lower quality, but pretty usable. Good for 4k context on 24GB.
3_5	3.5	6.0	21.5 GB	23.0 GB	25.0 GB	Lower quality, only use if you need more context on 24GB.
3_0	3.0	6.0	18.9 GB	20.4 GB	22.4 GB	Very low quality, pushes context to max but likely unusable.

Download instructions

With git:

git clone --single-branch --branch 6_5 https://huggingface.co/bartowski/Cerebrum-1.0-8x7b-exl2

With huggingface hub (credit to TheBloke for instructions):

pip3 install huggingface-hub

To download the main (only useful if you only care about measurement.json) branch to a folder called Cerebrum-1.0-8x7b-exl2:

mkdir Cerebrum-1.0-8x7b-exl2
huggingface-cli download bartowski/Cerebrum-1.0-8x7b-exl2 --local-dir Cerebrum-1.0-8x7b-exl2 --local-dir-use-symlinks False

To download from a different branch, add the --revision parameter:

Linux:

mkdir Cerebrum-1.0-8x7b-exl2-6_5
huggingface-cli download bartowski/Cerebrum-1.0-8x7b-exl2 --revision 6_5 --local-dir Cerebrum-1.0-8x7b-exl2-6_5 --local-dir-use-symlinks False

Windows (which apparently doesn't like _ in folders sometimes?):

mkdir Cerebrum-1.0-8x7b-exl2-6.5
huggingface-cli download bartowski/Cerebrum-1.0-8x7b-exl2 --revision 6_5 --local-dir Cerebrum-1.0-8x7b-exl2-6.5 --local-dir-use-symlinks False