--- license: llama3.1 language: - en base_model: - nvidia/OpenMath2-Llama3.1-8B pipeline_tag: text-generation tags: - math - nvidia - llama --- ## GGUF quantized version of OpenMath2-Llama3.1-8B project original [source](https://huggingface.co/nvidia/OpenMath2-Llama3.1-8B) (base model) Q_2_K (not nice) Q_3_K_S (acceptable) Q_3_K_M is acceptable (good for running with CPU) Q_3_K_L (acceptable) Q_4_K_S (okay) Q_4_K_M is recommanded (balance) Q_5_K_S (good) Q_5_K_M (good in general) Q_6_K is good also; if you want a better result; take this one instead of Q_5_K_M Q_8_0 which is very good; need a reasonable size of RAM otherwise you might expect a long wait f16 is similar to the original hf model; opt this one or hf also fine; make sure you have a good machine ### how to run it use any connector for interacting with gguf; i.e., [gguf-connector](https://pypi.org/project/gguf-connector/) <style> .image-container { display: flex; justify-content: center; align-items: center; gap: 20px; } .image-container img { width: 350px; height: auto; } </style> <div class="image-container"> <img src="https://huggingface.co/nvidia/OpenMath2-Llama3.1-8B/resolve/main/scaling_plot.jpg" title="Performance of Llama-3.1-8B-Instruct as it is trained on increasing proportions of OpenMathInstruct-2"> <img src="https://huggingface.co/nvidia/OpenMath2-Llama3.1-8B/resolve/main/math_level_comp.jpg" title="Comparison of OpenMath2-Llama3.1-8B vs. Llama-3.1-8B-Instruct across MATH levels"> </div> the chart and figure above are from base model (nvidia side)