exl2 quants for ReMM V2.2
This repository includes the quantized models for the ReMM V2.2 model by Undi. ReMM is a model merge attempting to recreate MythoMax using the SLERP merging method and newer models.
Current models
exl2 Quant | Model Branch | Model Size | Minimum Recommended VRAM (4096 Context, fp16 cache) | BPW |
---|---|---|---|---|
3-Bit | main | 5.44 GB | 8GB GPU | 3.14 |
3-Bit | 3bit | 6.36 GB | 10GB GPU | 3.72 |
4-Bit | 4bit | 7.13 GB | 12GB GPU (10GB with swap) | 4.2 |
4-Bit | 4.6bit | 7.81 GB | 12GB GPU | 4.63 |
5-Bit | R136a1's Repo | 8.96 GB | 16GB GPU (12GB with swap) | 5.33 |
Where to use
There are a couple places you can use an exl2 model, here are a few:
- tabbyAPI
- Aphrodite Engine
- ExUI
- oobabooga's Text Gen Webui
- When using the downloader, make sure to format like this: Anthonyg5005/ReMM-v2.2-L2-13B-exl2:QuantBranch
- With 5-Bit download: R136a1/ReMM-v2.2-L2-13B-exl2
- KoboldAI (Clone repo, don't use snapshot)
How to download:
oobabooga's downloader
use something like download-model.py to download with python requests.
Install requirements:
pip install requests tqdm
Example for downloading 3bpw:
python download-model.py Anthonyg5005/ReMM-v2.2-L2-13B-exl2:3bit
huggingface-cli
You may also use huggingface-cli
To install it, install python hf-hub
pip install huggingface-hub
Example for 3bpw:
huggingface-cli download Anthonyg5005/ReMM-v2.2-L2-13B-exl2 --local-dir ReMM-v2.2-L2-13B-exl2-3bpw --revision 3bit
Git LFS (not recommended)
I would recommend the http downloaders over using git, they can resume downloads if failed and are much easier to work with.
Make sure to have git and git LFS installed.
Example for 3bpw download with git:
Have LFS file skip disabled
# windows
set GIT_LFS_SKIP_SMUDGE=0
# linux
export GIT_LFS_SKIP_SMUDGE=0
Clone repo branch
git clone https://huggingface.co/Anthonyg5005/ReMM-v2.2-L2-13B-exl2 -b 3bit
- Downloads last month
- 3
Inference API (serverless) does not yet support ExLlamaV2 models for this pipeline type.
Model tree for Anthonyg5005/ReMM-v2.2-L2-13B-exl2
Base model
Undi95/ReMM-v2.2-L2-13B
Finetuned
this model