You can deploy this model and start earning money today!

bartowski/
sophosympatheia_Glistening-Gem-31B-v2.0-GGUF

Name: sophosympatheia_Glistening-Gem-31B-v2.0-GGUF
Availability: InStock

gguf

mergekit

merge

not-for-all-audiences

text-generation

sophosympatheia/Glistening-Gem-31B-v2.0

Views

No views yet

Llamacpp imatrix Quantizations of Glistening-Gem-31B-v2.0 by sophosympatheia

Using llama.cpp release b10380 for quantization.

Original model: https://huggingface.co/sophosympatheia/Glistening-Gem-31B-v2.0

Model details:

Parameter count: 31B
Input support: text - mmproj from upstream was broken, see here
MTP: no
imatrix: yes - details

How to run

Prompt format

<bos><|turn>system
{system_prompt}<turn|>
<|turn>user
{prompt}<turn|>
<|turn>model
<|channel>thought
<channel|>

Don't know which to choose? Grab Q4_K_M (19.60GB) - usually a good mix of size and performance. Download instructions available here

Available files:

Filename	Quant type	File Size	Split	Description
sophosympatheia_Glistening-Gem-31B-v2.0-bf16.gguf	bf16	61.41GB	true	Full BF16 weights.
sophosympatheia_Glistening-Gem-31B-v2.0-Q8_0.gguf	Q8_0	32.64GB	false	Extremely high quality, generally unneeded but max available quant.
sophosympatheia_Glistening-Gem-31B-v2.0-Q6_K_L.gguf	Q6_K_L	27.07GB	false	Uses Q8_0 for embed and output weights. Very high quality, near perfect, recommended.
sophosympatheia_Glistening-Gem-31B-v2.0-Q6_K.gguf	Q6_K	26.73GB	false	Very high quality, near perfect, recommended.
sophosympatheia_Glistening-Gem-31B-v2.0-Q5_K_L.gguf	Q5_K_L	22.95GB	false	Uses Q8_0 for embed and output weights. High quality, recommended.
sophosympatheia_Glistening-Gem-31B-v2.0-Q5_K_M.gguf	Q5_K_M	22.61GB	false	High quality, recommended.
sophosympatheia_Glistening-Gem-31B-v2.0-Q5_K_S.gguf	Q5_K_S	21.50GB	false	High quality, recommended.
sophosympatheia_Glistening-Gem-31B-v2.0-Q4_K_L.gguf	Q4_K_L	19.94GB	false	Uses Q8_0 for embed and output weights. Good quality, recommended.
sophosympatheia_Glistening-Gem-31B-v2.0-Q4_1.gguf	Q4_1	19.77GB	false	Legacy format, similar performance to Q4_K_S but with improved tokens/watt on Apple silicon.
sophosympatheia_Glistening-Gem-31B-v2.0-Q4_K_M.gguf	Q4_K_M	19.60GB	false	Good quality, default size for most use cases, recommended.
sophosympatheia_Glistening-Gem-31B-v2.0-Q4_K_S.gguf	Q4_K_S	18.20GB	false	Slightly lower quality with more space savings, recommended.
sophosympatheia_Glistening-Gem-31B-v2.0-Q4_0.gguf	Q4_0	18.08GB	false	Legacy format, kept for compatibility with older tools.
sophosympatheia_Glistening-Gem-31B-v2.0-IQ4_NL.gguf	IQ4_NL	18.03GB	false	Similar to IQ4_XS, but slightly larger.
sophosympatheia_Glistening-Gem-31B-v2.0-IQ4_XS.gguf	IQ4_XS	17.16GB	false	Decent quality, smaller than Q4_K_S with similar performance, recommended.
sophosympatheia_Glistening-Gem-31B-v2.0-Q3_K_XL.gguf	Q3_K_XL	17.15GB	false	Uses Q8_0 for embed and output weights. Lower quality but usable, good for low RAM availability.
sophosympatheia_Glistening-Gem-31B-v2.0-Q3_K_L.gguf	Q3_K_L	16.81GB	false	Lower quality but usable, good for low RAM availability.
sophosympatheia_Glistening-Gem-31B-v2.0-Q3_K_M.gguf	Q3_K_M	15.92GB	false	Low quality.
sophosympatheia_Glistening-Gem-31B-v2.0-IQ3_M.gguf	IQ3_M	15.13GB	false	Medium-low quality, new method with decent performance comparable to Q3_K_M.
sophosympatheia_Glistening-Gem-31B-v2.0-Q3_K_S.gguf	Q3_K_S	14.33GB	false	Low quality, not recommended.
sophosympatheia_Glistening-Gem-31B-v2.0-IQ3_XS.gguf	IQ3_XS	13.84GB	false	Lower quality, new method with decent performance, slightly better than Q3_K_S.
sophosympatheia_Glistening-Gem-31B-v2.0-IQ3_XXS.gguf	IQ3_XXS	12.98GB	false	Lower quality, new method with decent performance, comparable to Q3 quants.
sophosympatheia_Glistening-Gem-31B-v2.0-Q2_K_L.gguf	Q2_K_L	12.97GB	false	Uses Q8_0 for embed and output weights. Very low quality but surprisingly usable.
sophosympatheia_Glistening-Gem-31B-v2.0-IQ2_M.gguf	IQ2_M	12.65GB	false	Relatively low quality, uses SOTA techniques to be surprisingly usable.
sophosympatheia_Glistening-Gem-31B-v2.0-Q2_K.gguf	Q2_K	12.63GB	false	Very low quality but surprisingly usable.
sophosympatheia_Glistening-Gem-31B-v2.0-IQ2_S.gguf	IQ2_S	12.08GB	false	Low quality, uses SOTA techniques to be usable.
sophosympatheia_Glistening-Gem-31B-v2.0-IQ2_XS.gguf	IQ2_XS	11.50GB	false	Low quality, uses SOTA techniques to be usable.
sophosympatheia_Glistening-Gem-31B-v2.0-IQ2_XXS.gguf	IQ2_XXS	10.83GB	false	Very low quality, uses SOTA techniques to be usable.

Download a specific file:

hf download bartowski/sophosympatheia_Glistening-Gem-31B-v2.0-GGUF --include "sophosympatheia_Glistening-Gem-31B-v2.0-Q4_K_M.gguf" --local-dir ./

Downloading using the Hugging Face CLI

Click to view download instructions

First, make sure you have the Hugging Face CLI installed:

pip install -U "huggingface_hub[cli]"

Download a specific file:

hf download bartowski/sophosympatheia_Glistening-Gem-31B-v2.0-GGUF --include "sophosympatheia_Glistening-Gem-31B-v2.0-Q4_K_M.gguf" --local-dir ./

The files marked true in the Split column above are stored as multiple parts in a folder. To download all the parts to a local folder, run:

hf download bartowski/sophosympatheia_Glistening-Gem-31B-v2.0-GGUF --include "sophosympatheia_Glistening-Gem-31B-v2.0-bf16/*" --local-dir ./

You can either specify a new local-dir (sophosympatheia_Glistening-Gem-31B-v2.0-bf16) or download them all in place (./)

How to run

These quants run with llama.cpp - installable in one line via llama.app:

curl -LsSf https://llama.app/install.sh | sh
llama-server -hf bartowski/sophosympatheia_Glistening-Gem-31B-v2.0-GGUF:Q4_K_M

llama-server includes a built-in chat web UI, served at http://localhost:8080 by default.

These quants were made with llama.cpp release b10380 - if this model's architecture is newly supported, you'll need that release or newer to run them.

They also work in: LM Studio · koboldcpp · ramalama · Jan AI · Text Generation Web UI · LoLLMs · Atomic Chat

imatrix

All quants made using imatrix option with dataset from here. The imatrix is available here: sophosympatheia_Glistening-Gem-31B-v2.0-imatrix.gguf.

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.

ARM/AVX information

llama.cpp automatically "repacks" weights into an interleaved layout at load time for faster inference on ARM and AVX machines - details in this PR. This once required downloading special Q4_0_4_4/4_8/8_8 files; those are long gone. Online repacking now covers Q4_0, IQ4_NL, and most K-quants, so no special quant choice is needed for CPU inference.

Which file should I choose?

Click here for details

An older (early 2024) but still useful write-up with charts comparing quant 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.

Hugging Face can also do this math for you: add your hardware in your Local Apps settings and the model page will show which files fit.

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, but will be slower than their K-quant equivalent, so speed vs performance is a tradeoff you'll have to decide.

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.

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