Exllama v2 Quantizations of Mistral-7B-Instruct-v0.2
Using turboderp's ExLlamaV2 experimental for quantization.
Each branch contains an individual bits per weight. This is an experimental ExLlamaV2 quantization with no measurement.json produced.
Default arguments used except when the bits per weight is above 6.0, at that point the lm_head layer is quantized at 8 bits per weight instead of the default 6.
Original model: https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.2
Download instructions
With git:
git clone --single-branch --branch 4_0 https://huggingface.co/bartowski/Mistral-7B-Instruct-v0.2-exl2
With huggingface hub (credit to TheBloke for instructions):
pip3 install huggingface-hub
To download a specific BPW branch to a folder called Mistral-7B-Instruct-v0.2-exl2:
mkdir Mistral-7B-Instruct-v0.2-exl2
huggingface-cli download bartowski/Mistral-7B-Instruct-v0.2-exl2 --revision 4_0 --local-dir Mistral-7B-Instruct-v0.2-exl2 --local-dir-use-symlinks False