Kai-fu Lee's AI company "Zero One Thousand Things" open source Yi large model was accused of plagiarism LLaMA

Li Kaifu, chairman and CEO of Innovation Works, founded the AI ​​large-model startup "Zero One Thousand Things" this year. The company has released two large open source models, Yi-34B and Yi-6B, and claims to be completely open to academic research and allows free commercial applications

On Hugging Face’s open source homepage, ehartford, the developer of the Yi-34B model, raised questions, saying that the model uses the Meta LLaMA architecture and only modified the names of two tensors (Tensor), namely input_layernorm and post_attention_layernorm

ehartford said that the open source community will definitely re-release the Yi large model and rename the tensors to conform to the LLaMA architecture. He hopes the company will consider implementing the change into the official model before it gains mass adoption, prompting a lot of chatter below.

In addition, IT Home found that a picture of Jia Yangqing, the former chief AI scientist of Alibaba, was circulated on the Internet today. The content stated that "a new model of a major domestic manufacturer is exactly the architecture of LLaMA, but in order to express the difference, the code is The name of the project was changed from LLaMA to their name, and then several variable names were changed."

According to reports from IT House, as of now, Zero One Thousand Things has not officially responded

Jia Yangqing announced his resignation from Alibaba in March this year and co-founded a new AI company, Lepton AI, in the United States, aiming to build an efficient AI application platform.

The above is the detailed content of Kai-fu Lee's AI company "Zero One Thousand Things" open source Yi large model was accused of plagiarism LLaMA. For more information, please follow other related articles on the PHP Chinese website!