New AI models now available in Amazon SageMaker JumpStart
AWS has introduced three new AI models in Amazon SageMaker JumpStart, enhancing capabilities in semantic analysis, object detection, and language processing.
AWS has announced that three new AI models are now accessible through Amazon SageMaker JumpStart: paraphrase-multilingual-MiniLM-L12-v2, Microsoft Table Transformer Detection, and Bielik-11B-v3.0-Instruct. These models are designed to enhance various AI applications by offering advanced capabilities in semantic analysis, object detection, and language processing.
The paraphrase-multilingual-MiniLM-L12-v2 model, developed by Sentence Transformers, provides a lightweight solution for semantic similarity tasks. It maps sentences and paragraphs into a 384-dimensional dense vector space, supporting over 50 languages. This model is particularly useful for applications such as cross-lingual semantic searches, multilingual document clustering, and sentence similarity scoring, without needing language-specific settings.
Microsoft’s Table Transformer Detection model is based on the DETR architecture and is specialized in identifying tables within unstructured documents like PDFs and scanned images. Trained on the PubTables-1M dataset, it is ideal for document digitization and automated data extraction processes, enabling efficient detection of tabular information in research papers, financial documents, and more.
The Bielik-11B-v3.0-Instruct model, created by SpeakLeash and ACK Cyfronet AGH, is a substantial generative language model featuring 11 billion parameters. It is trained on a multilingual dataset covering 32 European languages, with a strong focus on Polish. This model excels in Polish and European language dialogues, STEM and mathematical reasoning, logical tasks, and enterprise applications requiring sophisticated language understanding across European languages.
SageMaker JumpStart allows users to deploy these models easily with just a few clicks to meet their specific AI needs. Users can start using these models by accessing the Models section in SageMaker Studio or by employing the SageMaker Python SDK to deploy them to their AWS accounts. For further details on deploying and utilizing these foundation models, refer to the Amazon SageMaker JumpStart documentation.