1. [Topic](/topics/) [Vision AI](/metropolis) NVIDIA TAO # NVIDIA TAO NVIDIA TAO is an advanced suite of skills and tools for post-training vision language and vision foundation models. These post-training tools can now be used as agentic workflows with natural language prompts. Coding agents use these tools and skills to autonomously hit model accuracy targets by evaluating performance, determining the precise training data needed, and then mining existing data or synthetically generating needed data. This dramatically reduces the time and data needed to build high-performing AI solutions that are ready for deployment from the edge to the cloud. Applicable across a wide variety of industries, TAO excels at delivering custom industrial AI models for visual inspection, quality control, and robotic guidance. [Download TAO](https://github.com/NVIDIA-TAO/tao-skills-bank "Download from NGC")[Get Started ](https://docs.nvidia.com/tao/tao-toolkit/latest/text/getting_started.html "Quick Start Guide")[Forum](https://forums.developer.nvidia.com/c/accelerated-computing/intelligent-video-analytics/tao-toolkit/17 "TAO Forums") * * * ## How TAO Works NVIDIA TAO agent skills and tools let you go from model post training to production deployment seamlessly with coding agents. The process begins with selecting a pretrained foundation model from the TAO model zoo or bringing a third-party model with an architecture supported by TAO. Next, you can adapt the model to your domain by fine-tuning and optimizing it to be smaller and faster at runtime. Finally, the trained models can be exported into open formats for deployment across diverse environments—from edge to cloud—with NVIDIA DeepStream agent skills and tools. These fine-tuning agents skills ensure that high-performing vision AI models can be quickly customized, optimized, and deployed at scale faster than ever. ![What is TAO Toolkit and how does it fit into AI model development workflow?](https://developer.download.nvidia.com/images/tao/26cptx-robotics-diagram-tao-agentic-fine-tuning-ari.png) ### Get Started Learn how to get started with TAO agent skills. [Watch the Video](https://www.youtube.com/watch?v=9AQkVbx3fKA)[Explore all TAO Skills](https://github.com/NVIDIA-TAO/tao-skills-bank) ### TAO Documentation Browse documentation and learn how to get started on TAO. [See the Quick Start Guide](https://docs.nvidia.com/tao/tao-toolkit/latest/text/getting_started.html) ### TAO 7 Release Note Learn the new features released in the latest TAO 7. [Read the Release Note](https://docs.nvidia.com/tao/tao-toolkit/text/release_notes.html#version-list) * * * ## Key Features ### Accelerate Development With Agent Skills and Tools Seamlessly use TAO 7 fine-tuning agent skills and tools with any coding agent to post-train open models such as NVIDIA Cosmos™ 3 and VisualChangeNet with natural language prompts. ### Optimize Model Accuracy With Data Sourcing Skills Reduce development and training time with TAO agent skills for data enhanced fine-tuning. Automatically analyze model accuracy weaknesses with root-cause reasoning to determine the right data strategy, then fill the gap with real or synthetically generated data. ### Achieve High Accuracy With Automated Training Accelerate model fine-tuning and ship faster with automated hyperparameter optimization using the AutoML agent, which achieves the highest accuracy and converges quickly with LLM-generated reasoning. ### Reduce Data Preparation Time With Annotation Skills Manage, process, and prepare datasets for AI model training with agent skills that streamline the data pipeline process with tools for data ingestion, auto-labeling, and conversion to formats optimized for NVIDIA TAO. ### Run Efficiently, Deploy Anywhere TAO standardizes deployment with NVIDIA DeepStream skills for all supported models on edge or cloud. Use knowledge distillation and quantization to compress large models into efficient, edge-ready versions with minimal accuracy loss. * * * ## Models Explore a collection of vision language and vision foundation models that are compatible with TAO fine-tuning skills. ### Reasoning Vision Language Model Generate reasoning traces and customize reasoning VLMs with SFT and PEFT technique for deeper, domain-specific scene and context understanding. - [NVIDIA Cosmos 3](https://huggingface.co/collections/nvidia/cosmos3) ### Multimodal Embedding Models Use these models as pretrained starting points, making it easy to fine-tune models for domain-specific tasks for search and action embedding. - [SigLip2](https://catalog.ngc.nvidia.com/orgs/nvidia/teams/tao/models/siglip_v2?version=deployable_v1.1) - [C-RADIO / C-RADIO-CLIP](https://catalog.ngc.nvidia.com/orgs/nvidia/teams/tao/models/radio-clip?version=trainable_v1.0) - [Cosmos Embed](https://huggingface.co/nvidia/Cosmos-Embed1-448p-anomaly-detection) - [NV-DINOv2](https://catalog.ngc.nvidia.com/orgs/nvidia/teams/tao/models/nvdinov2_vitg?version=NV-DinoV2-518-V0) ### Grounding and Segmentation Models Easily combine pretrained vision models with a foundation model for tasks like detection, segmentation, classification, and change detection, streamlining domain-specific customization. - Real-time detection ([RT-DETR](https://docs.nvidia.com/tao/tao-toolkit/latest/text/cv_finetuning/pytorch/object_detection/rt_detr.html)) - Text-guided object segmentation ([MaskGroundingDINO](https://docs.nvidia.com/tao/tao-toolkit/latest/text/cv_finetuning/pytorch/instance_segmentation/mask_grounding_dino.html)) - Visual change detection ([Visual ChangeNet](https://docs.nvidia.com/tao/tao-toolkit/text/cv_finetuning/pytorch/visual_changenet/visual_changenet_classify.html#multiple-golden-data-format)) ### Depth Estimation Models Use mono and stereo depth estimation foundation models to achieve strong zero-shot generalization. - [FoundationStereo](https://catalog.ngc.nvidia.com/orgs/nvidia/teams/tao/models/foundationstereo?version=deployable_v2.0) - [Foundation Pose](https://catalog.ngc.nvidia.com/orgs/nvidia/teams/tao/models/foundationpose?version=deployable_v1.0) * * * ## Get Started With TAO ![icon representing workflows](https://developer.download.nvidia.com/icons/m48-interapp-workflow-complex.svg) ### Set Up Your System Check to see if your machine meets the system requirements and compatibility, then get started by installing TAO. [Hardware Requirement](https://docs.nvidia.com/tao/tao-toolkit/latest/text/getting_started.html#prerequisites) [Setup Options](https://docs.nvidia.com/tao/tao-toolkit/latest/text/getting_started.html#prerequisites) ![icon representing TAO resources](https://developer.download.nvidia.com/icons/m48-demo-topics.svg) ### TAO Github Tutorials and Notebooks Check out extended resources and Jupyter notebooks for TAO. [Learn More](https://github.com/NVIDIA-TAO/tao-tutorials) ![icon representing neural network](https://developer.download.nvidia.com/icons/m48-neural-network-3.svg) ### Download TAO Download TAO skills and find the latest TAO and models at NGC. [Download TAO Skills](https://github.com/NVIDIA-TAO/tao-skills-bank) * * * ## Starter Kits ### Agentic Post-Training of NVIDIA Cosmos 3 Seamlessly integrate the agent skills and tools with any AI coding agent to fine-tune open models including NVIDIA Cosmos 3 using natural language prompts. - GIthub: [Post-train Cosmos 3 on WTS with TAO agent skills](https://github.com/NVIDIA-TAO/tao-tutorials/blob/main/tutorials/tao_agent_skills_examples/post_train_cosmos3/post_train_cosmos3_lora.md) - Tech Blog: [Post-Train NVIDIA Cosmos 3 in One Day Using Agent Skills](https://developer.nvidia.com/blog/post-train-nvidia-cosmos-3-in-one-day-using-agent-skills/) - Tutorial: [Post-Train NVIDIA Cosmos 3 With TAO 7 Agent Skills](https://www.youtube.com/watch?v=9AQkVbx3fKA) - Livestream: [Post-Train NVIDIA Cosmos 3 In a Day with NVIDIA TAO Agent Skills](https://www.youtube.com/watch?v=lKaqbYa0HUE) ### Fine-Tuning Take advantage of Supervised Fine-Tuning (SFT) with labeled data and Self-Supervised Learning (SSL)​ with unlabeled data. - [Chain-of-Thought Reasoning Auto-Labeling](https://github.com/NVIDIA/skills/blob/main/skills/tao-train-mask-auto-label/SKILL.md) ### Advanced Fine-Tuning Take advantage of various fine-tuning methodologies based on your training data strategy, including Supervised Fine-Tuning (SFT) with labeled data and Self-Supervised Learning (SSL)​ with unlabeled data. - [Data Efficient Fine-Tuning](https://github.com/NVIDIA/skills/tree/main/skills/tao-run-deft-aoi) - [Self-Supervised Learning](https://docs.nvidia.com/tao/tao-toolkit/text/cv_finetuning/pytorch/self_supervised_learning/index.html) - [Computer Vision Fine-Tuning](https://docs.nvidia.com/tao/tao-toolkit/text/cv_finetuning/index.html) * * * ## General FAQ NVIDIA TAO is free to download. Use of this software is governed by the [NVIDIA Software License Agreement](https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-software-license-agreement/) and [Product-Specific Terms](https://www.nvidia.com/en-us/agreements/enterprise-software/product-specific-terms-for-ai-products/) for NVIDIA AI Products., which allows training and deploying AI models, including for commercial use. No. TAO is designed to simplify AI model customization with pre-trained models, automated workflows, and low-code tools. With TAO agentic skills, now your coding agents can use TAO to fine-tune models. Yes. TAO enables you to fine-tune NVIDIA pre-trained models using your own datasets to create solutions tailored to your specific use case. TAO can be used across industries including manufacturing, retail, healthcare, transportation, smart cities, and robotics to build customized vision AI applications. TAO-trained models can be deployed across cloud, data center, edge, and embedded environments using NVIDIA deployment platforms and tools. ## More Resources ![Decorative image representing Community](https://developer.download.nvidia.com/icons/m48-people-group.svg) ### Join the Community ![Decorative image representing Developer Newsletter](https://developer.download.nvidia.com/icons/m48-email-settings.svg) ### Sign Up for the Developer Newsletter ![Decorative image representing Developer Program](https://developer.download.nvidia.com/icons/m48-developer-1.svg) ### Join the NVIDIA Developer Program * * * ## Ethical AI NVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications. When downloaded or used in accordance with our terms of service, developers should work with their supporting model team to ensure this model meets requirements for the relevant industry and use case and addresses unforeseen product misuse. For more detailed information on ethical considerations for this model, please see the Model Card++ Explainability, Bias, Safety & Security, and Privacy Subcards. Please report security vulnerabilities or NVIDIA AI Concerns [here](https://www.nvidia.com/en-us/support/submit-security-vulnerability/). Get Started Today. [Start With NVIDIA TAO](https://docs.nvidia.com/tao/tao-toolkit/latest/text/getting_started.html)