Description
Training and evaluating LLMs needs data, while real data raises privacy issues and costly labeling. NeMo Data Designer is NVIDIA's synthetic data tool for generating high-quality datasets from scratch or from seed data.
Define fields, distributions and dependencies, generate text, code and multimodal data with LLMs, and validate quality automatically.
From scratch:Schema-defined data.
Seed data:Expand samples.
Dependencies:Consistent fields.
Validation:Automatic checks.
Define fields, distributions and dependencies, generate text, code and multimodal data with LLMs, and validate quality automatically.
Features
From scratch:Schema-defined data.
Seed data:Expand samples.
Dependencies:Consistent fields.
Validation:Automatic checks.
Screenshots
Tags:llm