File size: 2,351 Bytes
0403961 6e4085c 0403961 6e4085c 0403961 4410f3e 0403961 6e4085c |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 |
---
license: apache-2.0
datasets:
- POLARIS-Project/Polaris-Dataset-53K
base_model:
- POLARIS-Project/Polaris-4B-Preview
library_name: llama.cpp
tags:
- polaris
- gguf
- ollama
---
# Polaris GGUF Quants
This repository contains GGUF (GPT-Generated Unified Format) model files for the [Polaris](https://huggingface.co/POLARIS-Project/Polaris-4B-Preview) model.
These files were quantized using an importance matrix generated from the Polaris dataset for optimal performance. The imatrix file can be viewed on the Github for this repository (coming soon).
The importance matrix was generated using 10k semi-randomly chosen examples from the [Polaris Dataset](https://huggingface.co/datasets/POLARIS-Project/Polaris-Dataset-53K)
# Associated Github
The [Github](https://github.com/dzur658/Polaris-4B-Preview-IQ-GGUF/tree/main) for this project contains the required modelfile settings necessary to run this model with Ollama along with some of the scripts I used to create these
imatrix quantizations.
# Accreditation
All credit for the original Polaris model belongs to their team and their affiliated organizations. Thank you for all of your hard work, and open publishing of the models
and research 😊, the POLARIS recipe will be absolutely crucial for edge LLM computing and if you haven't read their [original paper](https://honorable-payment-890.notion.site/POLARIS-A-POst-training-recipe-for-scaling-reinforcement-Learning-on-Advanced-ReasonIng-modelS-1dfa954ff7c38094923ec7772bf447a1#20da954ff7c380289bf4e4b10d72b374) I'd reccomend giving it a good read to truly
appreciate it.
```
@misc{Polaris2025,
title = {POLARIS: A Post-Training Recipe for Scaling Reinforcement Learning on Advanced Reasoning Models},
url = {https://hkunlp.github.io/blog/2025/Polaris},
author = {An, Chenxin and Xie, Zhihui and Li, Xiaonan and Li, Lei and Zhang, Jun and Gong, Shansan and Zhong, Ming and Xu, Jingjing and Qiu, Xipeng and Wang, Mingxuan and Kong, Lingpeng}
year = {2025}
}
```
# Ollama Integration
To use these models with Ollama, llmstudio, llama.cpp, etc. select your desired quantization level from the dropdown menu (e.g., `Q4_K_M`) and use the provided command.
**Note:** Replace `Q4_K_M` with the tag for the specific model version you wish to download.
```bash
ollama pull hf.co/dzur658/Polaris-4B-Preview-IQ-GGUF:Q4_K_M |