> For clean Markdown of any page, append .md to the page URL. > For a complete documentation index, see https://docs.nvidia.com/nemo-helix/v0.8.0/documentation/customizer-reference/llms.txt. > For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.nvidia.com/nemo-helix/_mcp/server. # About Fine-Tuning Learn how to fine-tune models by making requests to NVIDIA NeMo Customizer 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. [Create a Model Entity](/documentation/customizer-reference/manage-model-entities/overview) pointing to your base model checkpoint (stored as a FileSet). 2. Format a compatible [dataset](/documentation/customizer-reference/tutorials/format-training-dataset). 3. [Create a customization job](/documentation/customizer-reference/manage-customization-jobs) referencing the Model Entity. 4. Monitor the job until it completes. 5. The customization job automatically creates either: * **LoRA jobs**: An adapter attached to the original Model Entity * **Full fine-tuning jobs**: A new Model Entity with the customized weights 1. [Deploy the model](/documentation/models-and-inference) using the Deployment Management Service. 2. Move on to [Evaluate the output model](/documentation/evaluate-models). --- ## Container Images Fine-tuning jobs run in container images published to NVIDIA NGC. Use the image tag that matches your NeMo Helix deployment release: | Image | Purpose | | -------------------------------------------------------------- | ----------------------------------------------------------------------------------------- | | `nvcr.io/nvidia/nemo-helix/nhx-customizer-tasks:` | Shared CPU task steps (file I/O, model entity, model spec) for all customization backends | | `nvcr.io/nvidia/nemo-helix/nhx-automodel-training:` | Automodel GPU training step (Docker) | | `nvcr.io/nvidia/nemo-helix/nhx-unsloth-training:` | Unsloth GPU training step (Docker) | | `nvcr.io/nvidia/nemo-helix/nhx-rl-training:` | RL (DPO and GRPO) GPU training and Gym sandbox on Kubernetes/Ray | These public images can be pulled directly from `nvcr.io`: ```bash docker pull nvcr.io/nvidia/nemo-helix/nhx-customizer-tasks: docker pull nvcr.io/nvidia/nemo-helix/nhx-automodel-training: docker pull nvcr.io/nvidia/nemo-helix/nhx-unsloth-training: docker pull nvcr.io/nvidia/nemo-helix/nhx-rl-training: ``` Most users do not need to pull these images manually. NeMo Helix resolves them from the configured platform image registry and tag. Pull or mirror them when you are operating a self-managed deployment, preloading an air-gapped environment, or validating registry access. Automodel and Unsloth images are used by the Docker job backend. For air-gapped or registry-restricted deployments, preload or mirror `nhx-rl-training` on the Kubernetes cluster that runs Ray. If your environment requires authenticated registry pulls, authenticate to `nvcr.io` with an NGC API key before pulling or mirroring the images. --- ## Model Catalog Explore the model families and sizes supported by NVIDIA NeMo Customizer. #### [Nemotron Models](/documentation/customizer-reference/models/llama-nemotron) View the available Nemotron models from NVIDIA, including Nemotron 3.5 Lightning, Nemotron 3 Nano and Super, and Llama Nemotron variants. #### [Llama Models](/documentation/customizer-reference/models/llama) View the available Llama models in the model catalog. #### [Phi Models](/documentation/customizer-reference/models/phi) View the available Phi models from Microsoft, designed for strong reasoning capabilities with efficient deployment. #### [GPT-OSS Models](/documentation/customizer-reference/models/gpt-oss) View the available GPT-OSS models supported for Full SFT and LoRA customization. #### [Embedding Models](/documentation/customizer-reference/models/embedding) View the available embedding models for question-answering and retrieval tasks. ## Task Guides Perform common fine-tuning tasks. #### [Manage Customization Jobs](/documentation/customizer-reference/manage-customization-jobs) Create, list, view, and cancel customization jobs. #### [Manage Model Entities](/documentation/customizer-reference/manage-model-entities/overview) Create FileSets and Model Entities to prepare base models for customization. #### [Manage Datasets](/documentation/get-started/core-concepts/manage-files) Upload and manage datasets for training. --- ## Tutorials Follow these tutorials to learn how to accomplish common fine-tuning tasks. #### [Format Training Datasets](/documentation/customizer-reference/tutorials/format-training-dataset) Learn how to format datasets for different model types. datasets chat-models completion-models #### [Start a LoRA Customization Job](/documentation/customizer-reference/tutorials/lora-customization-job) Learn how to start a LoRA customization job using a custom dataset. nemo-customizer #### [Start a Full SFT Customization Job](/documentation/customizer-reference/tutorials/sft-customization-job) Learn how to start an SFT customization job using a custom dataset. nemo-customizer #### [Start a DPO Customization Job](/documentation/customizer-reference/tutorials/dpo-customization-job) Learn how to start a DPO customization job using preference data. nemo-customizer dpo #### [Distill a Model with Knowledge Distillation](/documentation/customizer-reference/tutorials/distillation-customization-job) Learn how to compress a larger teacher model into a smaller student model. nemo-customizer knowledge-distillation #### [Check Customization Job Metrics](/documentation/customizer-reference/tutorials/metrics) Learn how to check job metrics using MLFlow or Weights & Biases. nemo-customizer mlflow wandb #### [Optimize Tokens per GPU](/documentation/customizer-reference/tutorials/optimize-throughput) Learn how to optimize the token-per-GPU throughput for a LoRA optimization job. nemo-customizer wandb sequence-packing --- ## References #### [Hyperparameters](/documentation/customizer-reference/manage-customization-jobs/training-configuration) View the available hyperparameters and their valid options that you can set when creating a customization job. #### [Customizer API](/documentation/reference/api-reference) View the OpenAPI specification for Customizer. #### [Troubleshoot Failed Jobs](/documentation/reference/troubleshooting/customizer) View troubleshooting tips for failed jobs. ## Docs - [Customization Concepts](https://docs.nvidia.com/nemo-helix/v0.8.0/documentation/customizer-reference/customization-concepts.md): - [GRPO and Reward Environments](https://docs.nvidia.com/nemo-helix/v0.8.0/documentation/customizer-reference/grpo-and-reward-environments.md): Train a model with GRPO against a NeMo Gym reward environment. - [Manage Customization Jobs](https://docs.nvidia.com/nemo-helix/v0.8.0/documentation/customizer-reference/manage-customization-jobs.md): - [Create Job](https://docs.nvidia.com/nemo-helix/v0.8.0/documentation/customizer-reference/manage-customization-jobs/create-a-customization-job.md): - [Get Job Status](https://docs.nvidia.com/nemo-helix/v0.8.0/documentation/customizer-reference/manage-customization-jobs/get-job-status.md): - [List Active Jobs](https://docs.nvidia.com/nemo-helix/v0.8.0/documentation/customizer-reference/manage-customization-jobs/list-active-jobs.md): - [Cancel Job](https://docs.nvidia.com/nemo-helix/v0.8.0/documentation/customizer-reference/manage-customization-jobs/cancel-job.md): - [Training Configuration](https://docs.nvidia.com/nemo-helix/v0.8.0/documentation/customizer-reference/manage-customization-jobs/training-configuration.md): - [Customization Job Reference](https://docs.nvidia.com/nemo-helix/v0.8.0/documentation/customizer-reference/manage-customization-jobs/customization-job-reference.md): - [Manage Model Entities for Customization](https://docs.nvidia.com/nemo-helix/v0.8.0/documentation/customizer-reference/manage-model-entities/overview.md): - [Create a Model Entity](https://docs.nvidia.com/nemo-helix/v0.8.0/documentation/customizer-reference/manage-model-entities/create-a-model-entity.md): - [Create a Model FileSet](https://docs.nvidia.com/nemo-helix/v0.8.0/documentation/customizer-reference/manage-model-entities/create-a-model-file-set.md): - [Model Catalog](https://docs.nvidia.com/nemo-helix/v0.8.0/documentation/customizer-reference/models/model-catalog.md): - [Dataset Format Requirements](https://docs.nvidia.com/nemo-helix/v0.8.0/documentation/customizer-reference/models/dataset-format.md): - [Nemotron Models](https://docs.nvidia.com/nemo-helix/v0.8.0/documentation/customizer-reference/models/llama-nemotron.md): - [Embedding Models](https://docs.nvidia.com/nemo-helix/v0.8.0/documentation/customizer-reference/models/embedding.md): - [GPT-OSS Models](https://docs.nvidia.com/nemo-helix/v0.8.0/documentation/customizer-reference/models/gpt-oss.md): - [Llama Models](https://docs.nvidia.com/nemo-helix/v0.8.0/documentation/customizer-reference/models/llama.md): - [Mistral Models](https://docs.nvidia.com/nemo-helix/v0.8.0/documentation/customizer-reference/models/mistral.md): - [Phi Models](https://docs.nvidia.com/nemo-helix/v0.8.0/documentation/customizer-reference/models/phi.md): - [Qwen Models](https://docs.nvidia.com/nemo-helix/v0.8.0/documentation/customizer-reference/models/qwen.md): - [Fine-Tuning Tutorials](https://docs.nvidia.com/nemo-helix/v0.8.0/documentation/customizer-reference/tutorials.md): - [Understanding NeMo Customizer: Models, Training, and Resources](https://docs.nvidia.com/nemo-helix/v0.8.0/documentation/customizer-reference/tutorials/understanding-models-and-training.md): - [Format Training Dataset](https://docs.nvidia.com/nemo-helix/v0.8.0/documentation/customizer-reference/tutorials/format-training-dataset.md): - [GRPO Environment Packages](https://docs.nvidia.com/nemo-helix/v0.8.0/documentation/customizer-reference/tutorials/grpo-environment-packages.md): Package native-v1, wheels-v1, and adapter-wheels-v1 environment FileSets for GRPO. - [Import and Fine-Tune Private Hugging Face Models](https://docs.nvidia.com/nemo-helix/v0.8.0/documentation/customizer-reference/tutorials/import-hugging-face-models.md): - [Full SFT Customization](https://docs.nvidia.com/nemo-helix/v0.8.0/documentation/customizer-reference/tutorials/sft-customization-job.md): - [DPO Customization](https://docs.nvidia.com/nemo-helix/v0.8.0/documentation/customizer-reference/tutorials/dpo-customization-job.md): - [LoRA Model Customization](https://docs.nvidia.com/nemo-helix/v0.8.0/documentation/customizer-reference/tutorials/lora-customization-job.md): - [Knowledge Distillation Customization](https://docs.nvidia.com/nemo-helix/v0.8.0/documentation/customizer-reference/tutorials/distillation-customization-job.md): - [Embedding Model Customization](https://docs.nvidia.com/nemo-helix/v0.8.0/documentation/customizer-reference/tutorials/embedding-customization-job.md): - [Train on SPECTER or Triplet JSONL](https://docs.nvidia.com/nemo-helix/v0.8.0/documentation/customizer-reference/tutorials/embedding-triplet-dataset.md): - [Checking Your Customization Job Metrics](https://docs.nvidia.com/nemo-helix/v0.8.0/documentation/customizer-reference/tutorials/metrics.md): - [Optimize for Tokens/GPU Throughput](https://docs.nvidia.com/nemo-helix/v0.8.0/documentation/customizer-reference/tutorials/optimize-throughput.md):