Data Services

TAO Data Services provides a comprehensive solution for managing, processing, and preparing datasets for AI model training. It streamlines the data pipeline process with tools for data ingestion, annotation, augmentation, and conversion to formats optimized for NVIDIA TAO Toolkit.

The Data Services tools are available under the Dataset section:

• Annotations

• Augmentation

• Auto-Label

• Data Analytics

• Data Mining

• Gap Analysis

Action Outputs

These TAO Data Services commands and subtasks write two files to the results directory of the run: experiment.yaml, which holds the final configuration of the run, including command-line overrides, and status.json, which holds the status log of the run in JSON Lines format. The following commands and subtasks write both files:

• annotations: convert, merge, slice, and sparse4d_prepare

• augmentation: generate

• auto_label: generate

• analytics: analyze and validate

• image: validate

The kpi_analyze subtask of analytics and the gap_analysis, tmm, and embedding commands do not write these two files.