I have Ari Ali ecces is the optimization frame of the pattern

Prune you havea European start that worked on the compression algorithms for the patterns ai, makes his / her optimization frame Open Source Thursday.

Prick ai is created a frame that applies several methods of efficiency, as the hunting, amount of well, amount distillation, to a model of data.

“We still set up and uploading compressed patterns, combination that compresses your model compressed after buying” CTO John Rachwan has told Ctopr

Especially the frame of the pruness if you can evaluate if there is a significant loss of quality after compressing a model and show earn you get.

“If I have to use a methirter, we are similar to how to abbate it makes standardized and wild transformers, and sharpeness, and added.

The major labs are already using various methods of compression already. For example, Open was reliabated in distillation to create faster versions of their flags models.

This is likely as Open I Open developed Gpt-4 Turbo, a faster version of the vpt-4. I’m Similar: The Flow.1-fast The image generation model is a distilled version of the fluux’s pattern.1 by black forest labs.

The distillation is a technique used to extract knowledge from a large model ai with a “teacher” model. Developers send requests to a teacher model and record the produce. Answers are sometimes compared to a dataset to see how precise are they. These used are then used to form the student model, that entered the teacher’s behavior.

“For great companies, what they are usually making this mysticity is usually based on a quantity method. For examples of amount:” Rachwan said. “But you can’t find a tool all of them, they do all you easy to use and combine together. And this is the great value that the bound is now.”

While the proo, you support any manners, from great spread patterns, discourse and focus of computer vision, society is more specified on generation models specifically.

Some of the existing prache users to be included Scenario and it Fothory. I am In addition to the open source edition, prawns a business that the company the company of Advanced optimization function including an optimization optimization.

“The most exciting function we are released soon will be a compression agent”, Rachwan said. “Basically you give your pattern, say,” I want more speed but don’t leave my accuracy by more than 2% ‘. And then the agent just make magic. He’s found the best combination for you, return for you. You don’t have to do anything like a developer. “

Trial you charges from the hour for their version pro. “It’s similar to how to think of a GPU when you rent a GPU on AWS or any cloud service”, Rachwan said.

And if your model is a critical part of your infrastructure you will end a handful of money on the inference with the optimized model. For example, prunes aia has made a luxe pattern eight times younger without too loss using their compression frame. Petuna ai hoping his clients think about their compression frame as a paying investment.

Re-handed the 6.5 million funding of $ 6.5 investors in the startup includes adventures Eqt, Daphni, Daphni, DiPhi Motier and Kima Ventures.