A lightweight, data-driven frontend visualization for the Small Language Model (SLM) implementation guide.
This repository provides an interactive, educational dashboard comparing SLM and LLM performance, hardware needs, pipeline phases, and integration patterns.
- SLM pipeline summary (data curation, model selection, fine-tuning, evaluation)
- Radar chart comparing 7B SLM vs 70B+ LLM across key metrics
- VRAM bar chart showing inference and QLoRA training requirements
- Plotly scatter plot modeling dataset size vs training time
- Architecture diagram for context engine + SLM inference + agent execution
- Scroll reveal animations and responsive grid layout for visual storytelling
- React 19
- Vite 8
- Chart.js (via CDN)
- Plotly.js (via CDN)
- ESLint (setup with React Hooks rules)
src/App.jsx: main page, chart setup and pipeline contentsrc/main.jsx: renders the appsrc/index.css/src/App.css: styling and layoutpublic/index.html: plugin scripts for Chart.js + Plotly, fonts
# install
npm install
# run development server
npm run dev
# build production
npm run build
# preview production build
npm run preview
# lint all files
npm run lintOpen http://localhost:5173 in browser.
package.jsonincludes dependencies:react,react-dom@vitejs/plugin-react,viteeslint,@eslint/js,eslint-plugin-react-hooks,eslint-plugin-react-refresh
- Default React compiler is not enabled; this is fine for this learning/demo layout.
- Update chart values in
src/App.jsxunderradarLabelsRaw,barLabelsRaw, andplotlyData. - Adjust pipeline cards in the
section-pipelineJSX block. - Add new SLM content sections as needed in the JSX structure.
- Fork repository
- Create branch:
feature/<name> - Commit:
git commit -m "feat: <message>" - Push and open PR
MIT (or adapt to your preferred open-source license)