Spaces:
Running
Running
| title: README | |
| emoji: 🏢 | |
| colorFrom: green | |
| colorTo: blue | |
| sdk: static | |
| pinned: false | |
| license: mit | |
| # GA High School Sports Data Science | |
| Data journalist exploring AI applications in local sports coverage. Currently building fine-tuned language models for Georgia high school athletics using open-source tools and publicly available sports journalism. | |
| > This is a research space. | |
| ## Current Projects | |
| - 🏈 [georgia-sports-llama3-dpo](https://huggingface.co/kslote/georgia-sports-llama3-dpo) — Llama 3.1 8B fine-tuned with DPO on GHSA sports content | |
| - 📊 [georgia-high-school-sports](https://huggingface.co/datasets/kslote/georgia-high-school-sports) — 4,078 preference pairs for DPO training | |
| ## About This Work | |
| All models are trained on publicly available content from [GPB Sports](https://www.gpb.org/sports). Preference data was generated locally on Apple Silicon using [Ollama](https://ollama.com/) with Llama 3.1 8B and Mistral 7B. Fine-tuning was done with QLoRA on Google Colab. |