bartowski/Tess-v2.5-Phi-3-medium-128k-14B-GGUF-torrent

Llamacpp imatrix Quantizations of Tess-v2.5-Phi-3-medium-128k-14B

Using llama.cpp release b3152 for quantization.

Original model: https://huggingface.co/migtissera/Tess-v2.5-Phi-3-medium-128k-14B

All quants made using imatrix option with dataset from here

Prompt format

<|im_start|> system
{system_prompt}<|im_end|> 
<|im_start|> user
{prompt}<|im_end|> 
<|im_start|> assistant

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

FilenameQuant typeFile SizeDescription
Tess-v2.5-Phi-3-medium-128k-14B-Q8_0.ggufQ8_015.14GBExtremely high quality, generally unneeded but max available quant.
Tess-v2.5-Phi-3-medium-128k-14B-Q6_K.ggufQ6_K11.84GBVery high quality, near perfect, recommended.
Tess-v2.5-Phi-3-medium-128k-14B-Q5_K_M.ggufQ5_K_M10.48GBHigh quality, recommended.
Tess-v2.5-Phi-3-medium-128k-14B-Q5_K_S.ggufQ5_K_S10.03GBHigh quality, recommended.
Tess-v2.5-Phi-3-medium-128k-14B-Q4_K_M.ggufQ4_K_M8.99GBGood quality, uses about 4.83 bits per weight, recommended.
Tess-v2.5-Phi-3-medium-128k-14B-Q4_K_S.ggufQ4_K_S8.38GBSlightly lower quality with more space savings, recommended.
Tess-v2.5-Phi-3-medium-128k-14B-IQ4_XS.ggufIQ4_XS7.90GBDecent quality, smaller than Q4_K_S with similar performance, recommended.
Tess-v2.5-Phi-3-medium-128k-14B-Q3_K_L.ggufQ3_K_L7.94GBLower quality but usable, good for low RAM availability.
Tess-v2.5-Phi-3-medium-128k-14B-Q3_K_M.ggufQ3_K_M7.37GBEven lower quality.
Tess-v2.5-Phi-3-medium-128k-14B-IQ3_M.ggufIQ3_M6.92GBMedium-low quality, new method with decent performance comparable to Q3_K_M.
Tess-v2.5-Phi-3-medium-128k-14B-Q3_K_S.ggufQ3_K_S6.51GBLow quality, not recommended.
Tess-v2.5-Phi-3-medium-128k-14B-IQ3_XS.ggufIQ3_XS6.25GBLower quality, new method with decent performance, slightly better than Q3_K_S.
Tess-v2.5-Phi-3-medium-128k-14B-IQ3_XXS.ggufIQ3_XXS5.92GBLower quality, new method with decent performance, comparable to Q3 quants.
Tess-v2.5-Phi-3-medium-128k-14B-Q2_K.ggufQ2_K5.61GBVery low quality but surprisingly usable.
Tess-v2.5-Phi-3-medium-128k-14B-IQ2_M.ggufIQ2_M5.19GBVery low quality, uses SOTA techniques to also be surprisingly usable.
Tess-v2.5-Phi-3-medium-128k-14B-IQ2_S.ggufIQ2_S4.81GBVery low quality, uses SOTA techniques to be usable.
Tess-v2.5-Phi-3-medium-128k-14B-IQ2_XS.ggufIQ2_XS4.61GBVery 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/Tess-v2.5-Phi-3-medium-128k-14B-GGUF --include "Tess-v2.5-Phi-3-medium-128k-14B-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/Tess-v2.5-Phi-3-medium-128k-14B-GGUF --include "Tess-v2.5-Phi-3-medium-128k-14B-Q8_0.gguf/*" --local-dir Tess-v2.5-Phi-3-medium-128k-14B-Q8_0

You can either specify a new local-dir (Tess-v2.5-Phi-3-medium-128k-14B-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.