Yann LeCun Loves Kannada Llama

Meta AI chief Yann LeCun is quite impressed by the recently released Kannada Llama. “I love this. This is why open source AI platforms will win: it's the only way for AI to cater to highly diverse languages, cultures, values, and centers of interest.” he wrote on X, reposting Kannada Llama post.

Similarly, computer scientist Subbarao Kambhampati praised Kannada Llama and wrote, “Faster development of Indic LlaMAs--like this Kan-LLaMA by Tensonic--is my favorite upside of open-source LLMs! (It will take too long for big players to see much lucre in handling low resource languages otherwise!). Hoping to see a good Tel-LlaMA soon…”

Kannada Llama aka Kan-LLaMA, is a 7 billion Llama 2 model which is LoRA pre-trained and fine-tuned on “Kannada” token, built by a Mumbai-based company called Tensoic. It is built by  Adarsh Shirawalmath, a 2nd year B.Tech student at Vellore Institute of Technology. 

In an exclusive interview with AIM, Shirawalmath said that when the AWS Campus Fund was announced at VIT, he was very excited about securing funding for building AI models. However, the minimum requirement for that was having a registered company. “I said that if we’re planning on building this venture, let’s just do it,” he narrated, describing how they quickly registered a company within 15 days.

“We randomly came up with the name Tensoic, which means ‘Tensor’ plus ‘Logic’, and we had no goals then, we were just fishing stuff.” 

The company said it expanded Llama-2’s existing linguistic capabilities for Low Resource Indic languages and specifically Kannada by fine tuning on 600 Million Kannada tokens and subsequently fine-tuning on SOTA Instruction Datasets The company said it will release the models, code, datasets and the paper(eventually) under permissive licenses.

Besides Kannada Llama, several other Llamas based on Indic languages are emerging, including Tamil Llama, Odia Llama, and Telugu Llama.