Llamacpp imatrix Quantizations of calme-3.2-instruct-78b

Using llama.cpp release b4222 for quantization.

Original model: https://huggingface.co/MaziyarPanahi/calme-3.2-instruct-78b

All quants made using imatrix option with dataset from here

Run them in LM Studio

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:

Filename Quant type File Size Split Description
calme-3.2-instruct-78b-Q8_0.gguf Q8_0 82.85GB true Extremely high quality, generally unneeded but max available quant.
calme-3.2-instruct-78b-Q6_K.gguf Q6_K 69.01GB true Very high quality, near perfect, recommended.
calme-3.2-instruct-78b-Q5_K_M.gguf Q5_K_M 58.31GB true High quality, recommended.
calme-3.2-instruct-78b-Q5_K_S.gguf Q5_K_S 55.08GB true High quality, recommended.
calme-3.2-instruct-78b-Q4_K_M.gguf Q4_K_M 50.70GB true Good quality, default size for most use cases, recommended.
calme-3.2-instruct-78b-Q4_K_S.gguf Q4_K_S 46.95GB false Slightly lower quality with more space savings, recommended.
calme-3.2-instruct-78b-Q4_0.gguf Q4_0 44.34GB false Legacy format, offers online repacking for ARM CPU inference.
calme-3.2-instruct-78b-IQ4_NL.gguf IQ4_NL 44.28GB false Similar to IQ4_XS, but slightly larger. Offers online repacking for ARM CPU inference.
calme-3.2-instruct-78b-Q4_0_8_8.gguf Q4_0_8_8 44.19GB false Optimized for ARM and AVX inference. Requires 'sve' support for ARM (see details below). Don't use on Mac.
calme-3.2-instruct-78b-Q4_0_4_8.gguf Q4_0_4_8 44.19GB false Optimized for ARM inference. Requires 'i8mm' support (see details below). Don't use on Mac.
calme-3.2-instruct-78b-Q4_0_4_4.gguf Q4_0_4_4 44.19GB false Optimized for ARM inference. Should work well on all ARM chips, not for use with GPUs. Don't use on Mac.
calme-3.2-instruct-78b-Q3_K_XL.gguf Q3_K_XL 43.43GB false Uses Q8_0 for embed and output weights. Lower quality but usable, good for low RAM availability.
calme-3.2-instruct-78b-IQ4_XS.gguf IQ4_XS 42.56GB false Decent quality, smaller than Q4_K_S with similar performance, recommended.
calme-3.2-instruct-78b-Q3_K_L.gguf Q3_K_L 42.35GB false Lower quality but usable, good for low RAM availability.
calme-3.2-instruct-78b-Q3_K_M.gguf Q3_K_M 40.31GB false Low quality.
calme-3.2-instruct-78b-IQ3_M.gguf IQ3_M 37.93GB false Medium-low quality, new method with decent performance comparable to Q3_K_M.
calme-3.2-instruct-78b-Q3_K_S.gguf Q3_K_S 36.77GB false Low quality, not recommended.
calme-3.2-instruct-78b-IQ3_XXS.gguf IQ3_XXS 34.03GB false Lower quality, new method with decent performance, comparable to Q3 quants.
calme-3.2-instruct-78b-Q2_K_L.gguf Q2_K_L 33.06GB false Uses Q8_0 for embed and output weights. Very low quality but surprisingly usable.
calme-3.2-instruct-78b-Q2_K.gguf Q2_K 31.85GB false Very low quality but surprisingly usable.
calme-3.2-instruct-78b-IQ2_M.gguf IQ2_M 31.43GB false Relatively low quality, uses SOTA techniques to be surprisingly usable.
calme-3.2-instruct-78b-IQ2_XS.gguf IQ2_XS 28.99GB false Low quality, uses SOTA techniques to be usable.
calme-3.2-instruct-78b-IQ2_XXS.gguf IQ2_XXS 27.30GB false Very low quality, uses SOTA techniques to be usable.
calme-3.2-instruct-78b-IQ1_M.gguf IQ1_M 25.42GB 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.

Downloading using huggingface-cli

Click to view download instructions

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/calme-3.2-instruct-78b-GGUF --include "calme-3.2-instruct-78b-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/calme-3.2-instruct-78b-GGUF --include "calme-3.2-instruct-78b-Q8_0/*" --local-dir ./

You can either specify a new local-dir (calme-3.2-instruct-78b-Q8_0) or download them all in place (./)

Q4_0_X_X information

New: Thanks to efforts made to have online repacking of weights in this PR, you can now just use Q4_0 if your llama.cpp has been compiled for your ARM device.

Similarly, if you want to get slightly better performance, you can use IQ4_NL thanks to this PR which will also repack the weights for ARM, though only the 4_4 for now. The loading time may be slower but it will result in an overall speed incrase.

Click to view Q4_0_X_X information These are *NOT* for Metal (Apple) or GPU (nvidia/AMD/intel) offloading, only ARM chips (and certain AVX2/AVX512 CPUs).

If you're using an ARM chip, the Q4_0_X_X quants will have a substantial speedup. Check out Q4_0_4_4 speed comparisons on the original pull request

To check which one would work best for your ARM chip, you can check AArch64 SoC features (thanks EloyOn!).

If you're using a CPU that supports AVX2 or AVX512 (typically server CPUs and AMD's latest Zen5 CPUs) and are not offloading to a GPU, the Q4_0_8_8 may offer a nice speed as well:

Click to view benchmarks on an AVX2 system (EPYC7702)
model size params backend threads test t/s % (vs Q4_0)
qwen2 3B Q4_0 1.70 GiB 3.09 B CPU 64 pp512 204.03 ± 1.03 100%
qwen2 3B Q4_0 1.70 GiB 3.09 B CPU 64 pp1024 282.92 ± 0.19 100%
qwen2 3B Q4_0 1.70 GiB 3.09 B CPU 64 pp2048 259.49 ± 0.44 100%
qwen2 3B Q4_0 1.70 GiB 3.09 B CPU 64 tg128 39.12 ± 0.27 100%
qwen2 3B Q4_0 1.70 GiB 3.09 B CPU 64 tg256 39.31 ± 0.69 100%
qwen2 3B Q4_0 1.70 GiB 3.09 B CPU 64 tg512 40.52 ± 0.03 100%
qwen2 3B Q4_K_M 1.79 GiB 3.09 B CPU 64 pp512 301.02 ± 1.74 147%
qwen2 3B Q4_K_M 1.79 GiB 3.09 B CPU 64 pp1024 287.23 ± 0.20 101%
qwen2 3B Q4_K_M 1.79 GiB 3.09 B CPU 64 pp2048 262.77 ± 1.81 101%
qwen2 3B Q4_K_M 1.79 GiB 3.09 B CPU 64 tg128 18.80 ± 0.99 48%
qwen2 3B Q4_K_M 1.79 GiB 3.09 B CPU 64 tg256 24.46 ± 3.04 83%
qwen2 3B Q4_K_M 1.79 GiB 3.09 B CPU 64 tg512 36.32 ± 3.59 90%
qwen2 3B Q4_0_8_8 1.69 GiB 3.09 B CPU 64 pp512 271.71 ± 3.53 133%
qwen2 3B Q4_0_8_8 1.69 GiB 3.09 B CPU 64 pp1024 279.86 ± 45.63 100%
qwen2 3B Q4_0_8_8 1.69 GiB 3.09 B CPU 64 pp2048 320.77 ± 5.00 124%
qwen2 3B Q4_0_8_8 1.69 GiB 3.09 B CPU 64 tg128 43.51 ± 0.05 111%
qwen2 3B Q4_0_8_8 1.69 GiB 3.09 B CPU 64 tg256 43.35 ± 0.09 110%
qwen2 3B Q4_0_8_8 1.69 GiB 3.09 B CPU 64 tg512 42.60 ± 0.31 105%

Q4_0_8_8 offers a nice bump to prompt processing and a small bump to text generation

Which file should I choose?

Click here for details

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.

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

Downloads last month
4,216
GGUF
Model size
78B params
Architecture
qwen2

1-bit

2-bit

3-bit

4-bit

5-bit

6-bit

8-bit

Inference Examples
Inference API (serverless) has been turned off for this model.

Model tree for bartowski/calme-3.2-instruct-78b-GGUF

Quantized
(5)
this model