bartowski/xLAM-8x22b-r-GGUF

Llamacpp imatrix Quantizations of xLAM-8x22b-r

Using llama.cpp release b3634 for quantization.

Original model: https://huggingface.co/Salesforce/xLAM-8x22b-r

All quants made using imatrix option with dataset from here

Run them in LM Studio

Prompt format

<s>[INST] {prompt}[/INST] 

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

Filename	Quant type	File Size	Split	Description
xLAM-8x22b-r-Q8_0.gguf	Q8_0	149.43GB	true	Extremely high quality, generally unneeded but max available quant.
xLAM-8x22b-r-Q6_K.gguf	Q6_K	115.54GB	true	Very high quality, near perfect, recommended.
xLAM-8x22b-r-Q5_K_M.gguf	Q5_K_M	99.98GB	true	High quality, recommended.
xLAM-8x22b-r-Q4_K_M.gguf	Q4_K_M	85.60GB	true	Good quality, default size for must use cases, recommended.
xLAM-8x22b-r-Q4_K_S.gguf	Q4_K_S	80.49GB	true	Slightly lower quality with more space savings, recommended.
xLAM-8x22b-r-Q4_0.gguf	Q4_0	79.87GB	true	Legacy format, generally not worth using over similarly sized formats
xLAM-8x22b-r-Q4_0_8_8.gguf	Q4_0_8_8	79.52GB	true	Optimized for ARM and CPU inference, much faster than Q4_0 at similar quality.
xLAM-8x22b-r-Q4_0_4_4.gguf	Q4_0_4_4	79.52GB	true	Optimized for ARM and CPU inference, much faster than Q4_0 at similar quality.
xLAM-8x22b-r-IQ4_XS.gguf	IQ4_XS	75.49GB	true	Decent quality, smaller than Q4_K_S with similar performance, recommended.
xLAM-8x22b-r-Q3_K_XL.gguf	Q3_K_XL	72.77GB	true	Uses Q8_0 for embed and output weights. Lower quality but usable, good for low RAM availability.
xLAM-8x22b-r-Q3_K_L.gguf	Q3_K_L	72.59GB	true	Lower quality but usable, good for low RAM availability.
xLAM-8x22b-r-Q3_K_M.gguf	Q3_K_M	67.80GB	true	Low quality.
xLAM-8x22b-r-IQ3_M.gguf	IQ3_M	64.50GB	true	Medium-low quality, new method with decent performance comparable to Q3_K_M.
xLAM-8x22b-r-Q3_K_S.gguf	Q3_K_S	61.51GB	true	Low quality, not recommended.
xLAM-8x22b-r-IQ3_XXS.gguf	IQ3_XXS	54.91GB	true	Lower quality, new method with decent performance, comparable to Q3 quants.
xLAM-8x22b-r-Q2_K_L.gguf	Q2_K_L	52.31GB	true	Uses Q8_0 for embed and output weights. Very low quality but surprisingly usable.
xLAM-8x22b-r-Q2_K.gguf	Q2_K	52.11GB	true	Very low quality but surprisingly usable.
xLAM-8x22b-r-IQ2_M.gguf	IQ2_M	46.72GB	false	Relatively low quality, uses SOTA techniques to be surprisingly usable.
xLAM-8x22b-r-IQ2_XS.gguf	IQ2_XS	42.01GB	false	Low quality, uses SOTA techniques to be usable.
xLAM-8x22b-r-IQ2_XXS.gguf	IQ2_XXS	37.89GB	false	Very low quality, uses SOTA techniques to be usable.
xLAM-8x22b-r-IQ1_M.gguf	IQ1_M	32.74GB	false	Extremely low quality, not recommended.

Embed/output weights

Some of these quants (Q3_K_XL, Q4_K_L etc) are the standard quantization method with the embeddings and output weights quantized to Q8_0 instead of what they would normally default to.

Some say that this improves the quality, others don't notice any difference. If you use these models PLEASE COMMENT with your findings. I would like feedback that these are actually used and useful so I don't keep uploading quants no one is using.

Thanks!

Credits

Thank you kalomaze and Dampf for assistance in creating the imatrix calibration dataset

Thank you ZeroWw for the inspiration to experiment with embed/output

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/xLAM-8x22b-r-GGUF --include "xLAM-8x22b-r-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/xLAM-8x22b-r-GGUF --include "xLAM-8x22b-r-Q8_0/*" --local-dir ./

You can either specify a new local-dir (xLAM-8x22b-r-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