TapTechNews May 23rd news, Amazon's cloud division announced that it has partnered with the artificial intelligence startup HuggingFace to run thousands of AI models at a lower cost on Amazon's custom computing chips.
HuggingFace, valued at 4.5 billion US dollars (TapTechNews note: currently about 32.625 billion Chinese yuan), has become the center for AI researchers and developers to share chatbots or other AI software, and has received support from companies such as Amazon, Google under Alphabet, and NVIDIA. It is the main platform for developers to obtain and adjust open-source AI models (such as MetaPlatforms' Llama3).
When developers adjust the open-source AI model, they usually hope to use the model to drive the software. Amazon and HuggingFace announced on Wednesday that they have cooperated to achieve this goal on a custom Amazon Web Services (AWS) chip called Inferentia2.
For us, efficiency is very important - ensuring that as many people as possible can run the model and can run in the most cost-effective way, said Jeff Boudier, head of product and growth at HuggingFace.
AWS hopes to attract more AI developers to use its cloud services to deliver AI. Although NVIDIA dominates the market for training models, AWS believes its chips can run those trained models at a lower cost.
Matt Wood, who is in charge of AWS's artificial intelligence products, said that you may train these models once a month, but may perform thousands of inferences on them every hour, which is where Inferentia2 really shines.