> 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-4-26B-A4B-it

> Fine-tune gemma-4-26B-A4B-it for image-text-to-text tasks with NeMo AutoModel using checked-in recipes and distributed settings.

[Gemma-4-26B-A4B-it](https://huggingface.co/google/gemma-4-26B-A4B-it) has a checked-in NeMo AutoModel recipe for image-text-to-text.
The Hugging Face configuration declares the `Gemma4ForConditionalGeneration` architecture.

## Fine-Tune Gemma-4-26B-A4B-it

Follow the [installation instructions](/get-started/installation), then review the checked-in recipe and
its declared topology below before choosing a local or cluster launcher.

```bash
uv run automodel examples/vlm_finetune/gemma4/gemma4_26b_a4b_moe.yaml \
  --nproc-per-node 8
```

## Choose a Workflow

| Goal                                         | Start Here                                                                                                                                            |
| -------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------- |
| Run the primary recipe for this model        | Use the [recipe configuration](https://github.com/NVIDIA-NeMo/Automodel/blob/main/examples/vlm_finetune/gemma4/gemma4_26b_a4b_moe.yaml).              |
| Try another checked-in recipe for this model | Use the [alternate recipe](https://github.com/NVIDIA-NeMo/Automodel/blob/main/examples/vlm_finetune/gemma4/gemma4_26b_a4b_moe_medpix_ep8cp2_4k.yaml). |

### Configuration

| Setting  | Configuration                                                                |
| -------- | ---------------------------------------------------------------------------- |
| Hardware | -                                                                            |
| Strategy | FSDP2; tp\_size=1, cp\_size=1, ep\_size=8                                    |
| Nodes    | 1                                                                            |
| Features | Checkpointing (enabled=true), Transformer Engine, DeepEP (dispatcher=deepep) |
| Advanced | torch.bfloat16, attention=te, linear=te, dispatcher=deepep                   |

### More Recipes for This Model

* [`gemma4_26b_a4b_moe_medpix_ep8cp2_4k.yaml`](https://github.com/NVIDIA-NeMo/Automodel/blob/main/examples/vlm_finetune/gemma4/gemma4_26b_a4b_moe_medpix_ep8cp2_4k.yaml)
* [`gemma4_26b_a4b_moe_mock.yaml`](https://github.com/NVIDIA-NeMo/Automodel/blob/main/examples/vlm_finetune/gemma4/gemma4_26b_a4b_moe_mock.yaml)
* [`gemma4_26b_a4b_moe_packing.yaml`](https://github.com/NVIDIA-NeMo/Automodel/blob/main/examples/vlm_finetune/gemma4/gemma4_26b_a4b_moe_packing.yaml)
* [`gemma4_26b_a4b_moe_peft.yaml`](https://github.com/NVIDIA-NeMo/Automodel/blob/main/examples/vlm_finetune/gemma4/gemma4_26b_a4b_moe_peft.yaml)

## Model Reference

### Model Architecture

| Property     | Value                            |
| ------------ | -------------------------------- |
| Task         | Image-text-to-text               |
| Architecture | `Gemma4ForConditionalGeneration` |

### Available Models

* [`google/gemma-4-26B-A4B-it`](https://huggingface.co/google/gemma-4-26B-A4B-it)

## Related Resources

* [Vision-language model fine-tuning guide](/recipes-e2e-examples/gemma-3-3n)