> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.nvidia.com/nemo/automodel/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.nvidia.com/nemo/automodel/_mcp/server.

# gemma-3-4b-it

> Fine-tune gemma-3-4b-it for image-text tasks with NeMo AutoModel using checked-in full-parameter, LoRA, and distributed recipes.

[gemma-3-4b-it](https://huggingface.co/google/gemma-3-4b-it) is a multimodal Gemma 3 checkpoint for image-text inputs. NeMo AutoModel provides full-parameter and Low-Rank Adaptation (LoRA) recipes for supervised fine-tuning.

Set up NeMo AutoModel with the [latest container](https://catalog.ngc.nvidia.com/orgs/nvidia/containers/nemo-automodel) or follow the [installation instructions](/get-started/installation).

## Fine-Tune gemma-3-4b-it

From the repository root, run:

```bash
uv run automodel examples/vlm_finetune/gemma3/gemma3_vl_4b_cord_v2.yaml --nproc-per-node 8
```

## Choose a Workflow

| Goal                         | Start Here                                                                                                                                                                    |
| ---------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Fine-tune on CORD-v2         | Use [gemma3\_vl\_4b\_cord\_v2.yaml](https://github.com/NVIDIA-NeMo/Automodel/blob/main/examples/vlm_finetune/gemma3/gemma3_vl_4b_cord_v2.yaml).                               |
| Fine-tune with LoRA          | Use [gemma3\_vl\_4b\_cord\_v2\_peft.yaml](https://github.com/NVIDIA-NeMo/Automodel/blob/main/examples/vlm_finetune/gemma3/gemma3_vl_4b_cord_v2_peft.yaml).                    |
| Fine-tune with Megatron FSDP | Use [gemma3\_vl\_4b\_cord\_v2\_megatron\_fsdp.yaml](https://github.com/NVIDIA-NeMo/Automodel/blob/main/examples/vlm_finetune/gemma3/gemma3_vl_4b_cord_v2_megatron_fsdp.yaml). |
| Fine-tune on MedPix-VQA      | Use [gemma3\_vl\_4b\_medpix.yaml](https://github.com/NVIDIA-NeMo/Automodel/blob/main/examples/vlm_finetune/gemma3/gemma3_vl_4b_medpix.yaml).                                  |

## Model Reference

### Model Architecture

| Property                  | Value                                                                 |
| ------------------------- | --------------------------------------------------------------------- |
| Task                      | Image-text-to-text                                                    |
| Hugging Face Architecture | `Gemma3ForConditionalGeneration`                                      |
| Checkpoint                | [`google/gemma-3-4b-it`](https://huggingface.co/google/gemma-3-4b-it) |

## Related Resources

* [Gemma 3 and Gemma 3n Fine-Tuning Guide](/recipes-e2e-examples/gemma-3-3n)