Gemma 2 2B synthetic intelligence (AI) mannequin was launched by Google DeepMind on Thursday. It has change into the newest addition to the Gemma 2 household of AI fashions and joins Gemma 2 27B and 9B fashions. Regardless of the light-weight measurement, the corporate claims that it outperformed GPT-3.5 fashions on the LMSYS Chatbot Area benchmark. Alongside, the tech large additionally launched ShieldGemma, a collection of classifier fashions to filter the enter and output of Gemma 2, and Gemma Scope, a analysis device that provides insights into how the AI mannequin features.
Gemma 2 2B AI Mannequin’s Options
In a weblog put up on Google for Builders, the corporate introduced Gemma 2 2B, which has change into the smallest language mannequin of the household. Launched as an on-device AI mannequin, the put up highlighted that regardless of its small parameter measurement, the output is considerably increased than its weight class because it was distilled from bigger fashions. Nevertheless, the tech large didn’t reveal which AI fashions have been used for its coaching.
Google additionally claimed that the Gemma 2 2B AI mannequin outperformed the GPT-3.5 fashions on the massive mannequin methods organisation (LMSYS) Chatbot Area Elo rating. The AI mannequin is alleged to have acquired a rating of 1126, whereas Mixtral 8x7b Instruct v0.1 mannequin scored 1114 and GPT-3.5 scored 1106.
The AI mannequin has additionally been optimised to run on a variety of {hardware}. For edge units and cloud-based deployment, it has been fine-tuned for Vertex AI and Google Kubernetes Engine (GKE). It is usually optimised for Nvidia TensorRT-LLM library and has been made obtainable as an Nvidia NIM. Additional, Gemma 2 2B additionally integrates with Keras, JAX, Hugging Face, and different main platforms.
Since it’s an open-source AI mannequin, the open weights could be downloaded from Google’s Hugging Face itemizing, Kaggle, or Vertex AI Mannequin Backyard. It can be tried out on the Google AI Studio.
ShieldGemma
Picture Credit score: Google
Aside from Gemma 2, ShieldGemma, a collection of security classifiers that may detect and take away dangerous content material in each the enter and output of the AI mannequin, was additionally launched. Google stated the system will give attention to hate speech, harassment, sexually express content material, and harmful content material.
Lastly, Gemma Scope, a analysis device for academicians and builders was additionally launched. The system makes use of sparse autoencoders (SAEs) to pinpoint particular components throughout the mannequin to focus on how the decision-making course of works and the way the structure operates.
