Llamacpp imatrix Quantizations of DeepSeek-Coder-V2-Lite-Instruct

Using llama.cpp release b3166 for quantization.

Original model: https://huggingface.co/deepseek-ai/DeepSeek-Coder-V2-Lite-Instruct

All quants made using imatrix option with dataset from here

Prompt format

<|begin▁of▁sentence|>{system_prompt}

User: {prompt}

Assistant: <|end▁of▁sentence|>Assistant:

Download a file (not the whole branch) from below:

Filename Quant type File Size Description
DeepSeek-Coder-V2-Lite-Instruct-Q8_0_L.gguf Q8_0_L 17.09GB Experimental, uses f16 for embed and output weights. Please provide any feedback of differences. Extremely high quality, generally unneeded but max available quant.
DeepSeek-Coder-V2-Lite-Instruct-Q8_0.gguf Q8_0 16.70GB Extremely high quality, generally unneeded but max available quant.
DeepSeek-Coder-V2-Lite-Instruct-Q6_K_L.gguf Q6_K_L 14.56GB Experimental, uses f16 for embed and output weights. Please provide any feedback of differences. Very high quality, near perfect, recommended.
DeepSeek-Coder-V2-Lite-Instruct-Q6_K.gguf Q6_K 14.06GB Very high quality, near perfect, recommended.
DeepSeek-Coder-V2-Lite-Instruct-Q5_K_L.gguf Q5_K_L 12.37GB Experimental, uses f16 for embed and output weights. Please provide any feedback of differences. High quality, recommended.
DeepSeek-Coder-V2-Lite-Instruct-Q5_K_M.gguf Q5_K_M 11.85GB High quality, recommended.
DeepSeek-Coder-V2-Lite-Instruct-Q5_K_S.gguf Q5_K_S 11.14GB High quality, recommended.
DeepSeek-Coder-V2-Lite-Instruct-Q4_K_L.gguf Q4_K_L 10.91GB Experimental, uses f16 for embed and output weights. Please provide any feedback of differences. Good quality, uses about 4.83 bits per weight, recommended.
DeepSeek-Coder-V2-Lite-Instruct-Q4_K_M.gguf Q4_K_M 10.36GB Good quality, uses about 4.83 bits per weight, recommended.
DeepSeek-Coder-V2-Lite-Instruct-Q4_K_S.gguf Q4_K_S 9.53GB Slightly lower quality with more space savings, recommended.
DeepSeek-Coder-V2-Lite-Instruct-IQ4_XS.gguf IQ4_XS 8.57GB Decent quality, smaller than Q4_K_S with similar performance, recommended.
DeepSeek-Coder-V2-Lite-Instruct-Q3_K_L.gguf Q3_K_L 8.45GB Lower quality but usable, good for low RAM availability.
DeepSeek-Coder-V2-Lite-Instruct-Q3_K_M.gguf Q3_K_M 8.12GB Even lower quality.
DeepSeek-Coder-V2-Lite-Instruct-IQ3_M.gguf IQ3_M 7.55GB Medium-low quality, new method with decent performance comparable to Q3_K_M.
DeepSeek-Coder-V2-Lite-Instruct-Q3_K_S.gguf Q3_K_S 7.48GB Low quality, not recommended.
DeepSeek-Coder-V2-Lite-Instruct-IQ3_XS.gguf IQ3_XS 7.12GB Lower quality, new method with decent performance, slightly better than Q3_K_S.
DeepSeek-Coder-V2-Lite-Instruct-IQ3_XXS.gguf IQ3_XXS 6.96GB Lower quality, new method with decent performance, comparable to Q3 quants.
DeepSeek-Coder-V2-Lite-Instruct-Q2_K.gguf Q2_K 6.43GB Very low quality but surprisingly usable.
DeepSeek-Coder-V2-Lite-Instruct-IQ2_M.gguf IQ2_M 6.32GB Very low quality, uses SOTA techniques to also be surprisingly usable.
DeepSeek-Coder-V2-Lite-Instruct-IQ2_S.gguf IQ2_S 6.00GB Very low quality, uses SOTA techniques to be usable.
DeepSeek-Coder-V2-Lite-Instruct-IQ2_XS.gguf IQ2_XS 5.96GB Very low quality, uses SOTA techniques to be usable.

Downloading using huggingface-cli

First, make sure you have hugginface-cli installed:

pip install -U "huggingface_hub[cli]"

Then, you can target the specific file you want:

huggingface-cli download bartowski/DeepSeek-Coder-V2-Lite-Instruct-GGUF --include "DeepSeek-Coder-V2-Lite-Instruct-Q4_K_M.gguf" --local-dir ./

If the model is bigger than 50GB, it will have been split into multiple files. In order to download them all to a local folder, run:

huggingface-cli download bartowski/DeepSeek-Coder-V2-Lite-Instruct-GGUF --include "DeepSeek-Coder-V2-Lite-Instruct-Q8_0.gguf/*" --local-dir DeepSeek-Coder-V2-Lite-Instruct-Q8_0

You can either specify a new local-dir (DeepSeek-Coder-V2-Lite-Instruct-Q8_0) or download them all in place (./)

Which file should I choose?

A great write up with charts showing various performances is provided by Artefact2 here

The first thing to figure out is how big a model you can run. To do this, you'll need to figure out how much RAM and/or VRAM you have.

If you want your model running as FAST as possible, you'll want to fit the whole thing on your GPU's VRAM. Aim for a quant with a file size 1-2GB smaller than your GPU's total VRAM.

If you want the absolute maximum quality, add both your system RAM and your GPU's VRAM together, then similarly grab a quant with a file size 1-2GB Smaller than that total.

Next, you'll need to decide if you want to use an 'I-quant' or a 'K-quant'.

If you don't want to think too much, grab one of the K-quants. These are in format 'QX_K_X', like Q5_K_M.

If you want to get more into the weeds, you can check out this extremely useful feature chart:

llama.cpp feature matrix

But basically, if you're aiming for below Q4, and you're running cuBLAS (Nvidia) or rocBLAS (AMD), you should look towards the I-quants. These are in format IQX_X, like IQ3_M. These are newer and offer better performance for their size.

These I-quants can also be used on CPU and Apple Metal, but will be slower than their K-quant equivalent, so speed vs performance is a tradeoff you'll have to decide.

The I-quants are not compatible with Vulcan, which is also AMD, so if you have an AMD card double check if you're using the rocBLAS build or the Vulcan build. At the time of writing this, LM Studio has a preview with ROCm support, and other inference engines have specific builds for ROCm.

Want to support my work? Visit my ko-fi page here: https://ko-fi.com/bartowski

Downloads last month
10,974
GGUF
Model size
15.7B params
Architecture
deepseek2

2-bit

3-bit

4-bit

5-bit

6-bit

8-bit

32-bit

Inference Examples
Unable to determine this model's library. Check the docs .