About Fine-Tuning#
Learn how to fine-tune a model by making requests to the NVIDIA NeMo Customizer microservice through the API. Fine-tuned models you have created can be deployed using NVIDIA NIMs.
Fine-Tuning Workflow#
At a high level, the fine-tuning workflow consists of the following steps:
Choose a model or create a new customization target and customization config to train a custom model.
Format a compatible dataset.
Monitor the job until it completes.
If training
all_weights, deploy the model using Deployment Management Service.Move on to Evaluate the output model.
Model Catalog#
Explore the model families and sizes supported by the NVIDIA NeMo Customizer microservice.
View the available Llama models in the model catalog.
View the available Llama Nemotron models from NVIDIA, including Nano and Super variants for efficient and advanced instruction tuning.
View the available Phi models from Microsoft, designed for strong reasoning capabilities with efficient deployment.
View the available GPT-OSS models supported for Full SFT customization.
View the available embedding models for question-answering and retrieval tasks.
For hardware compatibility, A100 configurations work with B200 GPUs. Refer to configuration management for details.
Task Guides#
Perform common fine-tuning tasks.
Create, list, view, and delete customization targets.
View available customization configurations to use when creating a customization job.
Create, list, view, and cancel customization jobs.
Tutorials#
Follow these tutorials to learn how to accomplish common fine-tuning tasks.
Learn how to format datasets for different model types.
Learn how to start a LoRA customization job using a custom dataset.
Learn how to start a SFT customization job using a custom dataset.
Learn how to start a Knowledge Distillation (KD) job using a teacher and student model.
Learn how to check job metrics using MLFlow or Weights & Biases.
Learn how to optimize the token-per-GPU throughput for a LoRA optimization job.
References#
View the available hyperparameters and their valid options that you can set when creating a customization job.
View the OpenAPI specification for Customizer.
View troubleshooting tips for failed jobs.