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:

  1. Format a dataset.

  2. Choose a model and compatible dataset.

  3. Optionally create a customization target for the model if it hasn’t already been created.

  4. Create a customization job.

  5. Monitor the job until it completes.

  6. Move on to Evaluate the output model.


Model Catalog#

Explore the model families and sizes supported by the NVIDIA NeMo Customizer microservice.

Llama Models

View the available Llama models in the model catalog.

Llama Models
Llama Nemotron Models

View the available Llama Nemotron models from NVIDIA, including Nano and Super variants for efficient and advanced instruction tuning.

Llama Nemotron Models
Phi Models

View the available Phi models from Microsoft, designed for strong reasoning capabilities with efficient deployment.

Phi Models

Task Guides#

Perform common fine-tuning tasks.

Manage Customization Targets

Create, list, view, and delete customization targets.

Manage Customization Targets
Manage Customization Configs

View available customization configurations to use when creating a customization job.

Manage Customization Configuration
Manage Customization Jobs

Create, list, view, and cancel customization jobs.

Manage Customization Jobs

Tutorials#

Follow these tutorials to learn how to accomplish common fine-tuning tasks.

Format Training Datasets

Learn how to format datasets for different model types.

Format Training Dataset
Start a LoRA Customization Job

Learn how to start a LoRA customization job using a custom dataset.

Start a LoRA Model Customization Job
Start a Full SFT Customization Job

Learn how to start a SFT customization job using a custom dataset.

Start a Full SFT Customization Job
Start a Knowledge Distillation Job

Learn how to start a Knowledge Distillation (KD) job using a teacher and student model.

Start a Knowledge Distillation (KD) Customization Job
Check Customization Job Metrics

Learn how to check job metrics using MLFlow or Weights & Biases.

Checking Your Customization Job Metrics
Optimize Tokens per GPU

Learn how to optimize the token-per-GPU throughput for a LoRA optimization job.

Optimize for Tokens/GPU Throughput

References#

Hyperparameters

View the available hyperparameters and their valid options that you can set when creating a customization job.

Hyperparameter Options
Customizer API

View the OpenAPI specification for Customizer.

Customizer API
Troubleshoot Failed Jobs

View troubleshooting tips for failed jobs.

Troubleshooting NeMo Customizer