Google has announced that its Gemma 2 series, which includes light and open models, will be available to researchers and developers via Vertex AI starting next month. This series includes a 27 billion parameter member as well as a surprising addition of a 9 billion parameter member. Gemma 2 was introduced in May as a successor to the 2 billion and 7 billion parameter models, aiming to run on Nvidia GPUs or a single TPU host on Vertex AI for developers looking to integrate AI into various devices.
The Gemma 2 models, like their predecessors, reflect the evolving AI landscape by offering smaller and lighter models to cater to different user needs. By providing options of 9 billion and 27 billion parameters, Google allows developers to choose how they want to use these models, either on the device or via the cloud. These open-source models can be easily customized and integrated into different projects.
In addition to the existing variants of Gemma, which include CodeGemma, Recurring Gemma, and PaliGemma, developers can benefit from the two Gemma 2 models. Google also plans to launch a 2.6 billion parameter model soon, aiming to bridge the gap between lightweight accessibility and powerful performance.
Developers can access Gemma 2 in Google AI Studio, where they can download model weights from sites like Kaggle and Hugging face. Researchers can use Gemma 2 for free through Kaggle or with a free tier for Colab notebooks.
For those interested in learning more about AI trends and networking with industry leaders, VentureBeat Transform 2024 will feature leaders from OpenAI, Chevron, Nvidia, Kaiser Permanente, and Capital One. This three-day event will provide essential insights on GenAI and offer opportunities to expand your professional network. Don’t miss out on this exclusive event happening in San Francisco from July 9-11.
Article Source
https://venturebeat.com/ai/googles-gemma-2-series-launches-with-not-one-but-two-lightweight-model-options-a-9b-and-27b/