Create README.md

README.md

@@ -0,0 +1,106 @@
+---
+license: other
+base_model: Skywork/Skywork-OR1-Math-7B
+tags:
+  - llama-cpp
+  - gguf
+  - qwen2
+  - math
+  - reasoning
+language:
+  - en
+pipeline_tag: text-generation
+library_name: gguf
+---
+
+# Skywork-OR1-Math-7B - GGUF
+
+GGUF quantizations of [Skywork/Skywork-OR1-Math-7B](https://huggingface.co/Skywork/Skywork-OR1-Math-7B).
+
+## Model Description
+
+Skywork-OR1-Math-7B is a math-specialized reasoning model from the Skywork Open Reasoner 1 (OR1) series, trained using large-scale rule-based reinforcement learning. It is fine-tuned on top of DeepSeek-R1-Distill-Qwen-7B with a curated dataset of 110K verifiable math problems and 14K coding questions.
+
+It achieves **69.8 on AIME24** and **52.3 on AIME25**, outperforming all similarly sized models on mathematical reasoning tasks.
+
+## Benchmark Results
+
+| Model | AIME24 (Avg@32) | AIME25 (Avg@32) |
+|---|---|---|
+| DeepSeek-R1-Distill-Qwen-7B | 55.5 | 39.2 |
+| Light-R1-7B-DS | 59.1 | 44.3 |
+| **Skywork-OR1-Math-7B** | **69.8** | **52.3** |
+
+## Available Quantizations
+
+| File | Bits | Size (approx) | Use Case |
+|---|---|---|---|
+| model_f16.gguf | 16 | ~15.3 GB | Maximum quality, reference |
+| model_q8_0.gguf | 8 | ~8.1 GB | Best quality, if VRAM allows |
+| model_q6_k.gguf | 6 | ~6.3 GB | Very high quality |
+| model_q5_k_m.gguf | 5 | ~5.5 GB | Recommended balance |
+| model_q5_k_s.gguf | 5 | ~5.3 GB | High quality, slightly smaller |
+| model_q4_k_m.gguf | 4 | ~4.7 GB | Good quality, widely used |
+| model_q4_k_s.gguf | 4 | ~4.5 GB | Smaller, slight quality loss |
+| model_q3_k_l.gguf | 3 | ~4.0 GB | Low VRAM, moderate quality |
+| model_q3_k_m.gguf | 3 | ~3.7 GB | Low VRAM |
+| model_q3_k_s.gguf | 3 | ~3.5 GB | Minimal VRAM |
+| model_q2_k.gguf | 2 | ~2.9 GB | Lowest quality, smallest size |
+
+## VRAM Requirements
+
+| Quantization | VRAM (approx) |
+|---|---|
+| F16 | ~16 GB |
+| Q8_0 | ~9 GB |
+| Q6_K | ~7 GB |
+| Q5_K_M | ~6 GB |
+| Q4_K_M | ~5.5 GB |
+| Q3_K_M | ~4.5 GB |
+| Q2_K | ~3.5 GB |
+
+## Usage
+
+### llama.cpp
+
+```bash
+./llama-cli -m model_q4_k_m.gguf -p "Solve: What is the sum of all prime numbers less than 20?" -n 512
+```
+
+### llama-cpp-python
+
+```python
+from llama_cpp import Llama
+
+llm = Llama(model_path="model_q4_k_m.gguf", n_ctx=4096)
+output = llm("Solve: What is the sum of all prime numbers less than 20?", max_tokens=512)
+print(output["choices"][0]["text"])
+```
+
+### LM Studio
+
+Download any `.gguf` file from this repository and load it directly in [LM Studio](https://lmstudio.ai/).
+
+### Ollama
+
+```bash
+ollama run hf.co/tinyopsec/Skywork-OR1-Math-7B-GGUF:Q4_K_M
+```
+
+## Original Model
+
+- **Original model:** [Skywork/Skywork-OR1-Math-7B](https://huggingface.co/Skywork/Skywork-OR1-Math-7B)
+- **Base model:** [deepseek-ai/DeepSeek-R1-Distill-Qwen-7B](https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Qwen-7B)
+- **Paper:** [Skywork Open Reasoner 1 Technical Report](https://arxiv.org/abs/2505.22312)
+- **License:** [Qwen License](https://huggingface.co/Qwen/Qwen1.5-7B/blob/main/LICENSE)
+
+## Citation
+
+```bibtex
+@article{he2025skywork,
+  title={Skywork Open Reasoner 1 Technical Report},
+  author={He, Jujie and Liu, Jiacai and Liu, Chris Yuhao and Yan, Rui and Wang, Chaojie and Cheng, Peng and Zhang, Xiaoyu and Zhang, Fuxiang and Xu, Jiacheng and Shen, Wei and Li, Siyuan and Zeng, Liang and Wei, Tianwen and Cheng, Cheng and An, Bo and Liu, Yang and Zhou, Yahui},
+  journal={arXiv preprint arXiv:2505.22312},
+  year={2025}
+}
+```