Google launches Gemma 2, its next generation of open models

Constructed for builders and researchers

Gemma 2 shouldn’t be solely extra highly effective, it is designed to extra simply combine into your workflows:

• Open and accessible: Identical to the unique Gemma fashions, Gemma 2 is offered underneath our commercially-friendly Gemma license, giving builders and researchers the flexibility to share and commercialize their improvements.
• Broad framework compatibility: Simply use Gemma 2 together with your most well-liked instruments and workflows due to its compatibility with main AI frameworks like Hugging Face Transformers, and JAX, PyTorch and TensorFlow through native Keras 3.0, vLLM, Gemma.cpp, Llama.cpp and Ollama. As well as, Gemma is optimized with NVIDIA TensorRT-LLM to run on NVIDIA- accelerated infrastructure or as an NVIDIA NIM inference microservice. You may fine-tune as we speak with Keras and Hugging Face. We’re actively working to allow further parameter-efficient fine-tuning choices.
• Easy deployment: Beginning subsequent month, Google Cloud clients will have the ability to simply deploy and handle Gemma 2 on Vertex AI.

Discover the brand new Gemma Cookbook, a set of sensible examples and recipes to information you thru constructing your personal functions and fine-tuning Gemma 2 fashions for particular duties. Uncover how one can simply use Gemma together with your tooling of alternative, together with for frequent duties like retrieval-augmented era.

Accountable AI improvement

We’re dedicated to offering builders and researchers with the sources they should construct and deploy AI responsibly, together with by way of our Responsible Generative AI Toolkit. The not too long ago open-sourced LLM Comparator helps builders and researchers with in-depth analysis of language fashions. Beginning as we speak, you should use the companion Python library to run comparative evaluations together with your mannequin and information, and visualize the ends in the app. Moreover, we’re actively engaged on open sourcing our textual content watermarking expertise, SynthID, for Gemma fashions.

When coaching Gemma 2, we adopted our sturdy inside security processes, filtering pre-training information and performing rigorous testing and analysis towards a complete set of metrics to establish and mitigate potential biases and dangers. We publish our outcomes on a big set of public benchmarks associated to security and representational harms.