Llama 2 on Vertex AI

Llama 2 on Vertex AI

Vertex AI has broadened its generative AI development capabilities with the introduction of new models. These models, now available in their Model Garden, further underline their commitment to cater to their customers by providing diverse options within an open ecosystem. Among the new additions are Llama 2 and Code Llama from Meta, as well as the Falcon LLM from the Technology Innovation Institute. Moreover, there’s an anticipation of Anthropic’s upcoming Claude 2. Such additions position Google Cloud with a carefully curated range of models spanning first-party, open source, and third-party options.

For many clients, the journey into generative AI starts within Vertex AI’s Model Garden. Here, they can access a vast array of carefully curated large models through APIs. This resource aids developers and data scientists in choosing the most suitable models for their particular needs. The selection can be based on various criteria such as capabilities, size, customization potential, and more. This ensures that users not only get powerful models but also enjoy the flexibility to adjust and deploy these models at a large scale.

Over 100 models in Model Garden, including Llama 2 and Claude 2

Now, Vertex AI offers even more options with the introduction of two significant new models: Meta’s Llama 2 and TII’s Falcon. Additionally, there’s a buzz about the soon-to-be-released Claude 2 from Anthropic. The diverse offerings of the Model Garden enable enterprises to align models to their unique requirements. Furthermore, when there’s a need for complete transparency into model details — especially for compliance and audit functions — open-source alternatives like Llama 2 and Falcon are ideal.

Recognizing the crucial role of an organization’s data in model performance, Vertex AI has simplified the model-tuning process. Notably, they stand out as the only cloud provider supporting Llama 2 with both adapter tuning and Reinforcement Learning with Human Feedback (RLHF). This provision allows organizations to adjust Llama 2 to their specific enterprise data, ensuring they retain complete control and ownership. Clients can undertake this tuning process in the recently unveiled Colab Enterprise, a fully managed data science notebook platform.

Read related articles: