| IBM Updates Granite Models |
| Written by Kay Ewbank |
| Monday, 28 October 2024 |
|
IBM has released new Granite models that it says provide state-of-the-art performance relative to model size. The Granite 3.0 collection includes a new, instruction-tuned, dense decoder-only LLM. The Granite series is a collection of generative AI models that apply generative AI to the modalities of language and code. The models come in different sizes and are built on a decoder-only architecture. They are trained on business-relevant domains, including financial, legal, IT, coding, and academic domains. The model type is designed to be used for retrieval augmented generation (RAG) for searching knowledge bases to generate tailored responses to customer inquiries, or to condense content into short descriptions.
IBM says the new Granite 3.0 8B Instruct has been trained using a novel two-phase method on over 12 trillion tokens of carefully vetted data across 12 different natural languages and 116 different programming languages. IBM says that fine-tuning smaller, fit-for-purpose models like Granite provides enterprises with a way to get frontier model performance at a fraction of the cost. Key to this is the ability to tailor Granite models to an organization's needs through InstructLab. InstructLab is an AI project developed by IBM and Red Hat that can be used by developers to enhance Large Language Models (LLMs) for specific business needs. It is open source and provides systematically generated synthetic data and phased-training protocols. All Granite models are released under the permissive Apache 2.0 license, and IBM is providing a detailed disclosure of training data sets and methodologies in the Granite 3.0 technical paper. IBM has also placed emphasis on model safety in this release, saying. Granite 3.0 8B Instruct demonstrates industry-leading robustness on the AttaQ benchmark, which measures an LLM's vulnerability to adversarial prompts designed to provoke models into generating harmful, inappropriate or otherwise undesirable prompts. The updated collection of Granite including recipes and how-to guides is available on Github, and developers can also experiment with the new Granite 3.0 8B Instruct model on the IBM Granite Playground. Granite 3.0 models are now available on IBM watsonx.ai through platform partners such as Google Vertex AI (through Google Cloud's Vertex AI Model Garden integrations with Hugging Face), Hugging Face, NVIDIA (as NIM microservices), Ollama and Replicate.
More InformationRelated ArticlesIBM Launches The Granite Code LLM Series IBM Releases Watsonx Granite Models IBM Announces WatsonX AI Platform To be informed about new articles on I Programmer, sign up for our weekly newsletter, subscribe to the RSS feed and follow us on Twitter, Facebook or Linkedin. 
Comments
or email your comment to: comments@i-programmer.info |
