Skip to main content
Ctrl+K
Tao Toolkit (Internal) - Home Tao Toolkit (Internal) - Home

Tao Toolkit (Internal)

Tao Toolkit (Internal) - Home Tao Toolkit (Internal) - Home

Tao Toolkit (Internal)

Table of Contents

Introduction

  • Overview

Migration Guides

  • Migrating to TAO 7.0
  • Migrating to TAO 6.0
  • Migrating to TAO 5.5
  • Migrating to the TAO API from TAO 5.3.0
  • Migrating from TAO 4.0.x to TAO 5.0.0
  • Migrating from TAO 3.x to TAO 4.0
  • Migrating from Legacy TLT to TAO

Getting Started

  • Getting Started with NVIDIA TAO Toolkit

Model Zoo

  • Overview
  • Foundation Models

Dataset

  • Data Annotation Format
  • Annotations
  • Offline Data Augmentation
  • Auto-Label
    • Mask Auto Label
    • Grounding DINO
    • 2D Grounding
    • Video Reasoning Annotation
  • Data Analytics
  • Data Mining
    • Generating Embeddings
    • Targeted Mode Matching (TMM)
  • Gap Analysis

Embedding Model Fine-tuning

  • Overview
  • CLIP
    • CLIP Introduction
    • CLIP Training, Evaluation, Inference, and Export
    • Using CLIP Embeddings
  • InternVideo2-CLIP
    • InternVideo2-CLIP Introduction
    • InternVideo2-CLIP Training, Evaluation, Inference, and Export
    • Using InternVideo2-CLIP Embeddings
  • Cosmos-Embed1

VLM Model Fine-tuning

  • Vision-Language Model (VLM) Fine-Tuning
    • Cosmos-Reason

Computer Vision

  • Overview
  • Self-Supervised Learning
    • Self-Supervised Learning
    • Nv-DINOv2
    • DINOv3
    • Masked Autoencoders (MAE)
  • Image Classification
  • Object Detection
    • Grounding DINO
    • DINO
    • Co-DETR
    • OCDNet
    • Deformable DETR
    • RT-DETR
  • Segmentation
    • SegFormer
    • Instance Segmentation
      • Data Input for Instance Segmentation
      • Mask2former
      • Mask Auto Labeler
      • Mask Grounding DINO
      • OneFormer
  • Visual ChangeNet
    • Visual ChangeNet-Segmentation
    • Visual ChangeNet-Classification
  • Depth Estimation - Monocular and Stereo
    • Monocular Depth Estimation
    • Stereo Depth Estimation
    • Fast Foundation Stereo
  • 3D Perception
    • PointPillars
    • BEVFusion
    • Sparse4D
      • Sparse4D
    • Panoptic 3D Reconstruction with NVIDIA TAO Toolkit
      • NvPanoptix3D
    • CenterPose
      • CenterPose
  • Recognition & Re-Identification
    • Metric Learning Recognition
      • Metric Learning Recognition
    • ReIdentificationNet
      • ReIdentificationNet
    • ReIdentificationNet Transformer
      • ReIdentificationNet Transformer
    • Character Recognition
      • OCRNet
    • Pose Classification
      • PoseClassificationNet
    • ActionRecognitionNet
  • Optical Inspection
    • SiameseOI

Model Optimization

  • Knowledge Distillation
  • Model Pruning
  • Quantization Aware Training
  • Post-Training Quantization
  • Quantizing a model in TAO (TAO Quant)
    • 1. Terminology
    • 2. Getting Started
    • 3. Choosing a Backend
    • 4. Configuration
    • 6. Skipping Layers From Quantization
    • 7. TorchAO Backend (Weight-Only PTQ)
    • 8. ModelOpt PyTorch Backend (Static PTQ)
    • 9. ModelOpt ONNX Backend (Static PTQ)
    • 10. API Reference
    • 11. Extending TAO Quant With a Custom Backend
    • 12. Limitations and Current Status

AutoML

  • AutoML

Reference Workflows

  • Reference Workflows
    • Data Enhanced Fine-tuning (DEFT)
      • DEFT for Automated Optical Inspection
      • DEFT for Object Detection

Training Pipeline Features

  • Automatic Mixed Precision
  • Deterministic Training and Reproducibility
  • Best Checkpoint Saving
  • Visualizing Training
  • MLOps
    • Overview
    • TAO WandB Integration
    • TAO Clearml Integration

Deploying to Inference SDKs

  • TAO Deploy Overview
    • CenterPose with TAO Deploy
    • CLIP with TAO Deploy
    • InternVideo2-CLIP with TAO Deploy
    • Classification (PyTorch) with TAO Deploy
    • Classification (TF1) with TAO Deploy
    • Classification (TF2) with TAO Deploy
    • Deformable DETR with TAO Deploy
    • DINO with TAO Deploy
    • Grounding DINO with TAO Deploy
    • DetectNet_v2 with TAO Deploy
    • DSSD with TAO Deploy
    • EfficientDet (TF1) with TAO Deploy
    • EfficientDet (TF2) with TAO Deploy
    • Faster RCNN with TAO Deploy
    • LPRNet with TAO Deploy
    • MAE with TAO Deploy
    • Mask RCNN with TAO Deploy
    • Mask2former with TAO Deploy
    • MLRecogNet with TAO Deploy
    • Monocular Depth with TAO Deploy
    • Multitask Image Classification with TAO Deploy
    • OCDNet with TAO Deploy
    • RetinaNet with TAO Deploy
    • RT-DETR with TAO Deploy
    • SSD with TAO Deploy
    • Segformer with TAO Deploy
    • UNet with TAO Deploy
    • YOLOv3 with TAO Deploy
    • YOLOv4 with TAO Deploy
    • YOLOv4-tiny with TAO Deploy
    • OCRNet with TAO Deploy
    • SiameseOI with TAO Deploy
    • Stereo Depth with TAO Deploy
    • VisualChangeNet-Classification with TAO Deploy
    • VisualChangeNet-Segmentation with TAO Deploy
    • Mask Grounding DINO with TAO Deploy
    • PointPillars with TAO Deploy
    • Integrating TAO Models with NSight DL Designer
  • Integrating TAO CV Models with Triton Inference Server
  • Optimizing and Profiling with TensorRT
    • TRTEXEC with ActionRecognitionNet
    • TRTEXEC with BodyPoseNet
    • TRTEXEC with CenterPose
    • TRTEXEC with CLIP
    • TRTEXEC with Classification TF1/TF2/PyT
    • TRTEXEC with Deformable-DETR
    • TRTEXEC with DetectNet-v2
    • TRTEXEC with DINO
    • TRTEXEC with DSSD
    • TRTEXEC with EfficientDet TF1/TF2
    • TRTEXEC with Facial Landmarks Estimation
    • TRTEXEC with Faster RCNN
    • TRTEXEC with Grounding DINO
    • TRTEXEC with LPRNet
    • TRTEXEC with MAE
    • TRTEXEC with Metric Learning Recognition
    • TRTEXEC with Mask RCNN
    • TRTEXEC with Multitask Classification
    • TRTEXEC with OCDNet
    • TRTEXEC with OCRNet
    • TRTEXEC with PointPillars
    • TRTEXEC with PoseClassificationNet
    • TRTEXEC with ReIdentificationNet
    • TRTEXEC with ReIdentificationNet Transformer
    • TRTEXEC with RetinaNet
    • TRTEXEC with RT-DETR
    • TRTEXEC with Segformer
    • TRTEXEC with SiameseOI
    • TRTEXEC with SSD
    • TRTEXEC with UNet
    • TRTEXEC with YOLO_v3
    • TRTEXEC with YOLO_v4
    • TRTEXEC with YOLO_v4_tiny
    • TRTEXEC with VisualChangeNet
    • TRTEXEC with Mask2former
    • TRTEXEC with Mask Grounding DINO
  • Integrating TAO Models into DeepStream
    • Deploying to DeepStream for Classification TF1/TF2/PyTorch
    • Deploying to DeepStream for Multitask Classification
    • Deploying to DeepStream for DetectNet_v2
    • Deploying to DeepStream for Deformable DETR
    • Deploying to DeepStream for DINO
    • Deploying to DeepStream for DSSD
    • Deploying to DeepStream for EfficientDet
    • Deploying to DeepStream for FasterRCNN
    • Deploying to DeepStream for RetinaNet
    • Deploying to DeepStream for SSD
    • Deploying to DeepStream for YOLOv3
    • Deploying to DeepStream for YOLOv4
    • Deploying to DeepStream for YOLOv4-tiny
    • Deploying to DeepStream for MaskRCNN
    • Deploying to Deepstream for UNet
    • Deploying to Deepstream for Segformer
    • Deploying nvOCDR to DeepStream

More Information

  • Release Notes
  • Frequently Asked Questions
  • Troubleshooting Guide
  • Tutorial Videos
    • Getting started with NVIDIA TAO
    • Create Custom Multi-Modal Fusion Models
    • Use Visual Prompt for In-Context Segmentation with NVIDIA TAO
    • Estimate and Track Object Poses with the NVIDIA TAO FoundationPose Model
    • Open Vocabulary Object Detection with NVIDIA Grounding-DINO
    • Use Text Prompts for Auto-Labeling with NVIDIA TAO
    • Visualize Model Training with TensorBoard
  • Support Information
  • Acknowledgements

Deprecations

  • Deprecations
    • TensorFlow 2.x [Deprecated]
      • Image Classification (TF2)
      • EfficientDet (TF2)
    • TensorFlow 1.x [Deprecated]
      • Body Pose Estimation
        • Body Pose Estimation
      • Character Recognition
        • LPRNet
      • Emotion Classification
      • Facial Landmarks Estimation
        • Facial Landmarks Estimation
      • Gaze Estimation
      • Gesture Recognition
      • HeartRate Estimation
      • Instance Segmentation
        • Data Input for Instance Segmentation
        • MaskRCNN
      • Image Classification (TF1)
      • Multitask Image Classification
      • Object Detection
        • DetectNet_v2
        • FasterRCNN
        • YOLOv3
        • YOLOv4
        • YOLOv4-tiny
        • SSD
        • DSSD
        • RetinaNet
        • EfficientDet (TF1)
      • Semantic Segmentation
        • UNET
    • Bring Your Own Model (BYOM)[DEPRECATED]
  • InternVideo2-CLIP

InternVideo2-CLIP#

  • InternVideo2-CLIP Introduction
  • InternVideo2-CLIP Training, Evaluation, Inference, and Export
  • Using InternVideo2-CLIP Embeddings

previous

Using CLIP Embeddings

next

InternVideo2-CLIP Introduction
NVIDIA NVIDIA
Privacy Policy | Manage My Privacy | Do Not Sell or Share My Data | Terms of Service | Accessibility | Corporate Policies | Product Security | Contact

Copyright © 2026, NVIDIA Corporation.

Last updated on Sep 18, 2026.