Pruna AI Open Source AI Model Optimization Framework to Help Efficient Compression

Pruna AI is a European startup focusing on the development of AI model compression algorithms. Recently, the company announced that it will open source its optimization framework to help developers compress and optimize AI models more efficiently. The framework combines a variety of methods such as caching, pruning, quantization and distillation to improve model performance while standardizing the storage, loading and evaluation process of compressed models.

Pruna AI's framework supports a variety of model types, including large language models, diffusion models, speech recognition and computer vision models, and currently focuses on the optimization of image and video generation models. Its services are already used by businesses such as Scenario and PhotoRoom. In addition to the open source version, Pruna AI also offers an enterprise version, which includes advanced optimization features and compression agents. Users only need to set speed and accuracy requirements, and the agents will automatically find the best compression combination.

Pruna AI charges an hourly fee, helping businesses save inference costs by optimizing their models. For example, the company successfully reduced the size of an Llama model by eight times, with almost no accuracy loss. Recently, Pruna AI completed a $6.5 million seed financing, with investors including EQT Ventures, Daphni, Motier Ventures and Kima Ventures.

Project address: https://github.com/PrunaAI/pruna