QUICKSTART_HF.md

Quick Reference: HuggingFace Deployment

🚀 Deploy in 5 Steps

1️⃣ Create Space

Go to https://huggingface.co/spaces → New Space

• SDK: Gradio
• Python: 3.10
• Public: Yes

2️⃣ Clone Repository

git clone https://huggingface.co/spaces/<username>/<space-name>
cd <space-name>

3️⃣ Copy Files

cp -r /path/to/GradCAMPlusPlus_SkinLesion/* .

4️⃣ Push to HF

git add .
git commit -m "Deploy: Multimodal Skin Lesion App"
git push

5️⃣ Wait for Build

Monitor at: https://huggingface.co/spaces/<username>/<space-name>

---

📚 Key Files

File	Purpose
app.py	Entry point
requirements.txt	Dependencies
spaces.yaml	HF config
README.md	Documentation
DEPLOYMENT.md	Detailed guide
DEPLOYMENT_CHECKLIST.md	Verification list
CONTRIBUTING.md	Contribution guide

---

✅ What Was Prepared

• App code with type hints
• Dependencies optimized
• Configuration files created
• Documentation complete
• Checklist provided
• Contributing guidelines added
• License included
• Examples & templates ready

---

🔗 URLs

• HF Spaces: https://huggingface.co/spaces
• This Project: https://huggingface.co/spaces/`/`
• Gradio Docs: https://gradio.app/docs

---

💡 Pro Tips

• Speed: Enable GPU in spaces.yaml for faster inference
• Cost: CPU-only mode works but is ~3x slower
• Testing: Run python app.py locally first
• Debugging: Check Logs tab on HF Spaces
• Updates: Push changes and Space rebuilds automatically

---

❓ Troubleshooting

Issue	Solution
Models not found	Verify data/weights/ in repo
Out of memory	Add GPU or reduce batch size
Build fails	Check Python 3.10 in spaces.yaml
Slow inference	Enable GPU in spaces.yaml

---

📊 Resources

• Size: ~600 MB (4 models + preprocessing)
• RAM: 2-3 GB minimum
• Time: 30-60s first run, 5-15s cached
• GPU: Optional (A10G recommended)

---

Status: ✅ DEPLOYMENT READY

Go forth and deploy! 🎉

For detailed instructions, see DEPLOYMENT.md or DEPLOYMENT_CHECKLIST.md