SemiKong semiconductor-specific LLM speeds chip design
SemiKong is an open-source large language model that has been designed to support the design and manufacturing chips by capturing industry-specific expertise.
The LLM can be downloaded for free and serves as a foundation for companies to build proprietary models tailored to their needs. SemiKong can be used to aid technology decisions and optimizations and to provide a Chatbot-style interface for workers.
The LLM is the result of a collaborative open-source project out of the AI Alliance with strong support from member Aitotronics. Other particpants included engineers from Japan’s Tokyo Electron and Vietnam’s FPT Software. The AI Alliance is an academic and industry body that includes most of the major players in AI, with the exception of Nvidia and OpenAI.
SemiKong is based on Meta’s Llama 3.1 and was created with support from Meta, AMD and IBM amongst others, with Aitomatic’s DXA system becoming the basis of SemiKong’s deployment.
“The joint effort employing an open source approach allowed the team to pool resources and expertise, accelerating development,” said Christopher Nguyen, CEO of Aitotronic, in a blog entry on Meta’s website. He added that SemiKong outperforms several state-of-the-art generic closed-source LLMs in understanding and generating semiconductor-related content.
According to the Meta blog, the expected benefits from using SemiKong include a 20 to 30 percent reduction in time-to-market for new chip designs and a 15 to 25 percent improvement in chip design first-time-right rates. SemiKong is also expected to speed up the onboarding of new semiconductor professionals by 40 to 50 percent.
Aitomatic Domain-Expert Agents (DXAs) use a neurosymbolic AI architecture, combining symbolic reasoning with neural networks. This approach is said to ensure deterministic decision-making. This differs from purely probabilistic models used elsewhere and allows DXAs to provide reliable, explainable results suitable for industrial applications.
The latest 70B SemiKong LLM can be downloaded from the SemiKong website
Related links and articles:
https://github.com/aitomatic/semikong?
News articles:
LLMs, autonomous AI agents speed up Nvidia chip design
Next-generation AI chip design exploration at IMEC UK
SemiFive, HyperAccel complete design of LLM processor
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
