PRUNA AI Open Open Framework

Pruna aiAnd it is a European startup company that works on pressure algorithms for artificial intelligence models, which does its improvement framework Open source Thursday.

Pruna AI creates a frame that applies many efficiency methods, such as cache, pruning, quantitative measurement, to a specific AI model.

“We also unify the saving and downloading compressed models, applying groups of these pressure methods, as well as evaluating your compressed model after its pressure,” said Puna Ai participant and CTO John Rachwan for Techcrunch.

In particular, the PRUNA AI’s action frame can assess whether there is a significant loss of quality after compressing the model and the performance gains you get.

He added: “If I want to use a metaphor, we look like how to embrace uniform transformers and distributors – how to contact them, how to save them, download them, etc. We do the same, but for the methods of competence.”

The large AI laboratories have already used different pressure methods. For example, Openai depends on distillation to create faster versions than its main models.

This is likely how Openai GPT-4 Turbo has developed, a faster version of GPT-4. Likewise, the Flux.1-SCHNELL Photo generation model is a futable Flux.1 fate version of Black Forest Labs.

Distillation is a technique used to extract knowledge from the great Amnesty International model with the “teacher student” model. The developers send requests to the teacher model and register the outputs. The answers are sometimes compared to a data set to see how accurately. Then these outputs are used to train a student model, which is trained to bring the teacher’s behavior to be rounded.

“For major companies, what they usually do is to build these things at home. What you can find in the open source world usually depends on one way. For example, let’s say one quantity of LLMS, or the temporary storage method is the same for proliferation models.” “But you cannot find a tool that all you collect, making it easy to use and collect it together. This is the great value that Pruna brings now.”

While Pruna Ai supports any kind of models, from large language models to proliferation models, speech models to text and computer vision models, the company focuses more specifically on photo and video generation models at the present time.

Some current PRUNA AI users include scenario and Optical hall. In addition to the open source version, PRUNA AI has an offer to the institution with advanced improvement features including the improvement agent.

“The most exciting advantage that we will soon will be will be a pressure agent,” Rachawan said. Basically, you give it your model, you say: “I want more speed, but I don’t drop my accuracy by more than 2 %. And then, the worker will find his magic.

PRUNA AI is accused of an hour for its professional release. “It is similar to how you will consider the graphics processing unit when renting the AWS graphics processing unit or any cloud service,” Rashwan said.

If your model is an important part of Amnesty International’s infrastructure, you will end up providing a lot of money to infer with the improvement model. For example, PRUNA AI made Llama eight times smaller without a great loss using its pressure frame. Pruna Ai hopes that its customers will think about its pressure as an investment that pays itself.

Pruna Ai raised seed financing of $ 6.5 million a few months ago. Among the investors at EQT Ventures, DAPHNI, Motier Ventures and Kima Ventures.