NVIDIA NeMo Customizer for Developers
NVIDIA NeMo? Customizer is a high-performance, scalable microservice that simplifies fine-tuning and alignment of generative AI models for domain-specific use cases. The microservice supports popular customization techniques such as low-rank adaptation (LoRA), p-tuning, and supervised fine-tuning (SFT), with continued integration of the latest customization techniques.
NeMo Customizer, along with the other NeMo microservices, enables developers to create data flywheels that continuously optimize generative AI agents, enhancing the overall experience for end users.
How NVIDIA NeMo Customizer Works
NeMo Customizer provides an easy-to-use API that lets you customize generative AI models. Simply provide the dataset, model name, hyperparameters, and type of customization in the API payload. NeMo Customizer will initiate a job to tune the model, resulting in a customized version for you.
The architecture diagram below illustrates the flow for using NeMo Customizer, which can launch multiple customization jobs. In the depicted scenario, you can utilize the NVIDIA NeMo Curator to gather high-quality data and NeMo Customizer to create two customization workflows: one for fine-tuning and one for alignment tuning. These outputs, along with NVIDIA NIM?, allow you to deploy a customized model tailored to your specific use case.
Introductory Blog
Read how NeMo Customizer simplifies the alignment and customization of generative AI models.
Tutorials
Explore tutorials designed to help you build custom generative AI models with the NeMo Customizer microservice.
GTC Session
Understand how NeMo Customizer, along with other NeMo microservices, facilitates the customization of generative AI models and supports ongoing performance evaluation, ensuring models remain relevant and effective over time.
Performance
NeMo Customizer uses several parallelism techniques to reduce the training time for large models with support for multi-GPU and multi-node infrastructure. These methods operate together to enhance the training process, ensuring optimal use of resources and improved training performance.
Experience 1.8x Faster Customization With NeMo Customizer
The benchmark represents customizing Llama-3-8Bon one 8xH100 80G SXM with sequence packing (4096 pack size, 0.9958 packing efficiency).
On: customized with NeMo Customizer.
Off: customized with leading market alternatives.
NVIDIA NeMo Customizer Learning Library
More Resources
Ethical AI
NVIDIA’s platforms and application frameworks enable developers to build a wide array of AI applications. Consider potential algorithmic bias when choosing or creating the models being deployed. Work with the model’s developer to ensure that it meets the requirements for the relevant industry and use case; that the necessary instruction and documentation are provided to understand error rates, confidence intervals, and results; and that the model is being used under the conditions and in the manner intended.
The NeMo Customizer microservice will be available soon. Sign up to get notified.