NetsPresso
NetsPresso is a hardware-aware AI model optimization platform. NetsPresso powers on-device AI across industries, and it's the ultimate platform for hardware-aware AI model development. Lightweight models of LLaMA and Vicuna enable efficient text generation. BK-SDM is a lightweight version of Stable Diffusion models. VLMs combine visual data with natural language understanding. NetsPresso resolves Cloud and server-based AI solutions-related issues, such as limited network, excessive cost, and privacy breaches. NetsPresso is an automatic model compression platform that downsizes computer vision models to a size small enough to be deployed independently on the smaller edge and low-specification devices. Optimization of target models being key, the platform combines a variety of compression methods which enables it to downsize AI models without causing performance degradation.
Learn more
Vicuna
Vicuna-13B is an open-source chatbot trained by fine-tuning LLaMA on user-shared conversations collected from ShareGPT. Preliminary evaluation using GPT-4 as a judge shows Vicuna-13B achieves more than 90%* quality of OpenAI ChatGPT and Google Bard while outperforming other models like LLaMA and Stanford Alpaca in more than 90%* of cases. The cost of training Vicuna-13B is around $300. The code and weights, along with an online demo, are publicly available for non-commercial use.
Learn more
LLaVA
LLaVA (Large Language-and-Vision Assistant) is an innovative multimodal model that integrates a vision encoder with the Vicuna language model to facilitate comprehensive visual and language understanding. Through end-to-end training, LLaVA exhibits impressive chat capabilities, emulating the multimodal functionalities of models like GPT-4. Notably, LLaVA-1.5 has achieved state-of-the-art performance across 11 benchmarks, utilizing publicly available data and completing training in approximately one day on a single 8-A100 node, surpassing methods that rely on billion-scale datasets. The development of LLaVA involved the creation of a multimodal instruction-following dataset, generated using language-only GPT-4. This dataset comprises 158,000 unique language-image instruction-following samples, including conversations, detailed descriptions, and complex reasoning tasks. This data has been instrumental in training LLaVA to perform a wide array of visual and language tasks effectively.
Learn more
Ango Hub
Ango Hub is a quality-focused, enterprise-ready data annotation platform for AI teams, available on cloud and on-premise. It supports computer vision, medical imaging, NLP, audio, video, and 3D point cloud annotation, powering use cases from autonomous driving and robotics to healthcare AI.
Built for AI fine-tuning, RLHF, LLM evaluation, and human-in-the-loop workflows, Ango Hub boosts throughput with automation, model-assisted pre-labeling, and customizable QA while maintaining accuracy. Features include centralized instructions, review pipelines, issue tracking, and consensus across up to 30 annotators. With nearly twenty labeling tools—such as rotated bounding boxes, label relations, nested conditional questions, and table-based labeling—it supports both simple and complex projects. It also enables annotation pipelines for chain-of-thought reasoning and next-gen LLM training and enterprise-grade security with HIPAA compliance, SOC 2 certification, and role-based access controls.
Learn more
