🚀 Llamacpp imatrix Quantizations of Jan-nano by Menlo
This project offers quantized versions of the Menlo/Jan-nano model using llama.cpp, providing various quantization types for different performance and quality requirements.
🚀 Quick Start
Using llama.cpp release b5627 for quantization. The original model can be found at https://huggingface.co/Menlo/Jan-nano.
You can run these quantized models in LM Studio or directly with llama.cpp, or any other llama.cpp based project.
✨ Features
- Multiple Quantization Types: Offers a wide range of quantization types, such as bf16, Q8_0, Q6_K_L, etc., to meet different performance and quality needs.
- Online Repacking: Some quantization types support online repacking of weights, which can improve performance on ARM and AVX machines.
- Download Flexibility: Allows downloading specific files or multiple split files using the huggingface-cli.
📦 Installation
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/Menlo_Jan-nano-GGUF --include "Menlo_Jan-nano-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/Menlo_Jan-nano-GGUF --include "Menlo_Jan-nano-Q8_0/*" --local-dir ./
You can either specify a new local-dir (Menlo_Jan-nano-Q8_0) or download them all in place (./)
💻 Usage Examples
Prompt format
<|im_start|>system
{system_prompt}<|im_end|>
<|im_start|>user
{prompt}<|im_end|>
<|im_start|>assistant
<think>
</think>
📚 Documentation
Download a file (not the whole branch) from below:
Property |
Details |
Filename |
Jan-nano-bf16.gguf, Jan-nano-Q8_0.gguf, etc. |
Quant type |
bf16, Q8_0, Q6_K_L, Q6_K, etc. |
File Size |
8.05GB, 4.28GB, 3.40GB, etc. |
Split |
false |
Description |
Full BF16 weights, Extremely high quality, generally unneeded but max available quant, etc. |
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
Previously, you would download Q4_0_4_4/4_8/8_8, and these would have their weights interleaved in memory in order to improve performance on ARM and AVX machines by loading up more data in one pass.
Now, however, there is something called "online repacking" for weights. details in this PR. If you use Q4_0 and your hardware would benefit from repacking weights, it will do it automatically on the fly.
As of llama.cpp build b4282 you will not be able to run the Q4_0_X_X files and will instead need to use Q4_0.
Additionally, if you want to get slightly better quality for , 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.
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, but will be slower than their K-quant equivalent, so speed vs performance is a tradeoff you'll have to decide.
🔧 Technical Details
Previously, weights of Q4_0_4_4/4_8/8_8 were interleaved in memory to improve performance on ARM and AVX machines. Now, "online repacking" for weights is available, which can automatically repack weights on the fly if using Q4_0 and the hardware benefits from it.
Benchmark on an AVX2 system (EPYC7702)
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
📄 License
This project is licensed under the apache-2.0 license.
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.
Thank you to LM Studio for sponsoring my work.
Want to support my work? Visit my ko-fi page here: https://ko-fi.com/bartowski