SunnyData

Curated library of 141 ready-to-run model scripts spanning 33 categories and multiple modalities. Use these assets to explore, prototype, and deploy NLP, vision, audio, tabular, and multimodal workloads with minimal setup.

🚀 Quick Start

• Clone or download the repository
• Navigate to a model directory, for example cd sunnysett_models/Marketing
• Run the desired script: python bert_base_uncased.py
• Review the console output for dependency installation and sample inference

📦 Repository Layout

Pre trained model dataset/
├── sunnysett_models/                # Primary library (33 categories / 141 scripts)
├── sunnysett_gcp_ready/             # Legacy 10-category bundle kept for reference
├── sunnysett_models_metadata.csv    # Auto-generated metadata (CSV)
├── sunnysett_models_metadata.json   # Auto-generated metadata (JSON)
├── sunnysett_models_summary.md      # Inventory statistics and category breakdown
├── add_specialized_models.py        # Generator for persona and domain bundles
├── analyze_and_organize.py          # Audit/cleanup helper for model scripts
└── analysis_model_summary.json      # Derived analytics powering the summary files

Category Highlights

• 33 categories covering industries, personas, deployment profiles, and core modalities
• 135 neural/deep learning models plus 6 classical or AutoML forecasters
• Coverage spans Text/NLP, Vision, Audio, Tabular/Time-Series, and Multimodal pipelines
• Full counts and modality breakdowns live in sunnysett_models_summary.md

📊 Inventory Snapshot

• Total Categories: 33
• Total Models: 141
• Neural / Deep Learning: 135
• Classical / AutoML: 6
• Top Modalities: Text/NLP (91), Vision (15), Audio (12), Tabular/Time-Series (12), Vision+Text (11)
• For a complete table with modality and type splits per category, open sunnysett_models_summary.md

🛠️ Running the Scripts

• Each script auto-installs required packages on first run (via pip)
• Sample inference calls live inside run_inference functions—edit the sample payloads to try your own data
• To execute an entire category on Linux/macOS:
  • cd sunnysett_models/<Category>
  • for script in *.py; do python "$script"; done
• On Windows PowerShell:
  • Get-ChildItem *.py | ForEach-Object { python $_ }

🗂️ Metadata & Tooling

• sunnysett_models_metadata.json / .csv: regenerated automatically from the codebase; reflects every script with category, pipeline, modality, and Hugging Face URL inference
• analysis_model_summary.json: consolidated analytics (counts, modality mixes, type splits) consumed by the documentation
• add_specialized_models.py: reproduces persona and deployment bundles if you need to regenerate scripts
• analyze_and_organize.py: scans the library, reports duplicates, and highlights any gaps relative to catalog CSVs

🔁 Adding or Updating Models

• Drop a new script in the relevant subdirectory under sunnysett_models/<Category>
• Follow the existing template: docstring metadata, install_dependencies, load_model, and run_inference
• Run python analyze_and_organize.py to verify classification and metadata coverage
• After changes, run python refresh_metadata.py to regenerate the summary, CSV, and JSON inventories

☁️ Deployment Notes

• Scripts are tested with Python 3.10+ and install dependencies on demand
• For GCP or other cloud VMs, copy either sunnysett_models/ (expanded library) or sunnysett_gcp_ready/ (compact legacy bundle)
• Ensure outbound internet access so Hugging Face models can download weights on first execution

🤝 Contributing

• Follow the scripting convention already in place (docstring, helper functions, clean logging)
• Update sunnysett_models_summary.md and metadata files by re-running the generation helper after adding assets
• Validate new models locally before proposing PRs or deploying to cloud environments

📄 License

This project is released under the MIT License. Individual model checkpoints may carry additional licensing requirements—consult each model’s Hugging Face page.

🆘 Getting Help

• Review the docstring in each script for usage instructions and dependency notes
• Verify network access and Hugging Face availability if downloads fail
• Open an issue or discussion thread with reproducible steps if you uncover defects

---

SunnySett AI Model Discovery Platform — rapid exploration of production-ready model starters.