scripts/README.md

Pre-Session Setup Scripts

These scripts help you prepare for the HuggingFace Enabling Sessions by pre-downloading models, converting slides, and testing the Spaces app.

Quick Start (All-in-One)

bash scripts/setup.sh

This runs all three scripts in sequence and prepares everything for your session.

---

Individual Scripts

1️⃣ Download Models Locally

Purpose: Pre-cache all HuggingFace models on your laptop so demos run instantly

File: scripts/download_models.py

Usage:

python3 scripts/download_models.py

What it does:

• Downloads 5 pre-trained models (~2-3 GB total)
• Stores in ~/.cache/huggingface/hub
• Takes 10-20 minutes depending on internet speed
• Models: Sentiment, NER, QA, Summarization, Semantic Similarity

Why: Models only need to download once. After this, they're cached forever.

---

2️⃣ Convert Markdown Slides to PowerPoint

Purpose: Convert markdown slides to editable PPTX format for your presentation

File: scripts/convert_slides_to_pptx.py

Requirements:

pip install python-pptx

Usage:

python3 scripts/convert_slides_to_pptx.py

What it does:

• Converts slides/SESSION1_SLIDES.md → slides/SESSION1_SLIDES.pptx
• Converts slides/SESSION2_SLIDES.md → slides/SESSION2_SLIDES.pptx
• Applies basic formatting (headers, text, colors)

Next steps:

• Open the PPTX files in PowerPoint or LibreOffice
• Customize colors, fonts, add speaker notes
• Add images or animations as desired

Note: The conversion is basic. You can enhance the PPTX afterwards in PowerPoint.

---

3️⃣ Pre-Warm Spaces App

Purpose: Test that the Spaces app is running and cache models on HF servers

File: scripts/prewarm_spaces.py

Usage:

python3 scripts/prewarm_spaces.py

What it does:

• Checks if Spaces app is online
• Waits for startup if needed (first-time builds take 2-3 min)
• Provides session readiness checklist
• Shows timing and performance info

Expected output:

✅ Spaces app is online
✓ Sentiment Analysis cached
✓ NER cached
✓ QA cached
✓ Summarization cached
✓ Semantic Similarity cached

---

Setup Timeline

Task	Time	Notes
Model Download	10-20 min	Run once, cached forever
Slide Conversion	1-2 min	One-time only
Spaces Pre-warm	2-5 min	Depends on app startup
Total	15-30 min	Do this 30 min before session

---

Session Day Checklist

• Run bash scripts/setup.sh (30 min before session)
• Open slides in PowerPoint/LibreOffice
• Test one demo on Spaces (click "Analyze Sentiment")
• Share Spaces URL with attendees: https://huggingface.co/spaces/Shouryahere/infy
• Open SPEAKER_NOTES.md for timing reference
• Screen share test (zoom, teams, etc)
• Audio test

---

Troubleshooting

"ModuleNotFoundError: No module named 'python_pptx'"

pip install python-pptx

Models still downloading during session?

• Run scripts/download_models.py earlier (it's idempotent, safe to run multiple times)
• Check internet speed: Large models (BART) can take 5+ min on slow networks

Spaces URL not responding?

• Check: https://huggingface.co/spaces/Shouryahere/infy
• If showing "Building", wait 2-3 minutes
• If showing red error, check Spaces build logs

Slide conversion looks poorly formatted?

• This script does basic conversion; use PowerPoint to enhance
• Recommended: Open PPTX and manually adjust formatting to match your company branding

---

Manual Alternative

If you prefer not to run scripts:

1. Download models: transformers-cli download-model distilbert-base-uncased-finetuned-sst-2-english (repeat for each model)
2. Convert slides: Use CloudConvert, Zamzar, or open markdown in LibreOffice Impress
3. Pre-warm Spaces: Just click each button once when you're ready

---

Advanced Options

Custom model downloads

Edit config.py to change which models are downloaded

Custom slide formatting

Edit scripts/convert_slides_to_pptx.py to adjust colors, fonts, sizes

Offline demo fallback

Pre-record a video of running the demos locally in case Spaces is slow

---

📝 Questions? See SPEAKER_NOTES.md for more context.

🚀 Ready to present!