Pruna AI makes AI models faster, cheaper, smaller, greener with the pruna package.
- It supports various models including CV, NLP, audio, graphs for predictive and generative AI.
- It supports various hardware including GPU, CPU, Edge.
- It supports various compression algortihms including quantization, pruning, distillation, caching, recovery, compilation that can be combined together.
- You can either play on your own with smash/compression configurations or let the smashing/compressing agent find the optimal configuration [Pro].
- You can evaluate reliable quality and efficiency metrics of your base vs smashed/compressed models. You can set it up in minutes and compress your first models in few lines of code!
You can smash your own models by installing pruna with:
pip install pruna
You can start with simple notebooks to experience efficiency gains with:
| Use Case | Free Notebooks |
|---|---|
| 3x Faster Stable Diffusion Models | ⏩ Smash for free |
| Making your LLMs 4x smaller | ⏩ Smash for free |
| Smash your model with a CPU only | ⏩ Smash for free |
| Transcribe 2 hours of audio in less than 2 minutes with Whisper | ⏩ Smash for free |
| 100% faster Whisper Transcription | ⏩ Smash for free |
| Run your Flux model without an A100 | ⏩ Smash for free |
| x2 smaller Sana in action | ⏩ Smash for free |
For more details about installation and tutorials, you can check the Pruna AI documentation.
