Just a few days after releasing Kan-Llama [ಕನ್-LLama], an Indic LLM based on Kannada token, the researchers have released a playground to test the model.
Tensoic, in partnership with E2E Networks, which is one the biggest providers of cloud GPUs in India, have released the playground which is hooked to NVIDIA A100 GPUs. The creators said that the Xylem.AI platform is used for inference, which is superfast and with mind blowing inference stack.
Read: Meet the Creator of Kannada Llama
The current version does not support multi turn conversations, but allows messing with the parameters.
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.
The pre-training process occurred on a solo NVIDIA A100 80GB instance, requiring approximately 50 hours and incurring an estimated cost of $170. The resultant LoRA adapter attained a size of around 1.1GB.
Currently, the model is built on top of Llama 2, but Adarsh Shirawalmath, the creator of Kannada Llama, told AIM that the team is also planning to make it on top of Mistral’s models, but the dataset is a little messy and not ready for Indic models yet. They are also planning to release Gujarati LLM soon.
The team is still working on the research paper as Shirawalmath said that he wants to make it perfect.
The post Tensoic Releases Playground for Kannada Llama on NVIDIA A100s appeared first on Analytics India Magazine.