---
license: apache-2.0
---
# Templates - Structural Control (FLUX.2-klein-base-4B)

This model is one of the open-source Diffusion Templates series models from [DiffSynth-Studio](https://github.com/modelscope/DiffSynth-Studio). It is a ControlNet control model capable of precisely guiding the spatial structure, object outlines, and perspective of generated images through an input reference image.

* Open-source code: [DiffSynth-Studio](https://github.com/modelscope/DiffSynth-Studio)
* Technical report: [arXiv](https://arxiv.org/abs/2604.24351)
* Project page: [GitHub](https://modelscope.github.io/diffusion-templates-web/)
* Documentation: [English Version](https://diffsynth-studio-doc.readthedocs.io/en/latest/Diffusion_Templates/Introducing_Diffusion_Templates.html)、[中文版](https://diffsynth-studio-doc.readthedocs.io/zh-cn/latest/Diffusion_Templates/Introducing_Diffusion_Templates.html)
* Online demo: [ModelScope](https://modelscope.cn/studios/DiffSynth-Studio/Diffusion-Templates)
* Models: [ModelScope](https://modelscope.cn/collections/DiffSynth-Studio/KleinBase4B-Templates)、[ModelScope International](https://modelscope.ai/collections/DiffSynth-Studio/KleinBase4B-Templates)、[HuggingFace](https://huggingface.co/collections/DiffSynth-Studio/kleinbase4b-templates)
* Datasets: [ModelScope](https://modelscope.cn/collections/DiffSynth-Studio/ImagePulseV2)、[ModelScope International](https://modelscope.cn/collections/DiffSynth-Studio/ImagePulseV2)、[HuggingFace](https://huggingface.co/collections/DiffSynth-Studio/imagepulsev2)

## Result Examples

|Condition|Prompt: A cat is sitting on a stone, bathed in bright sunshine.|Prompt: A cat is sitting on a stone, surrounded by colorful magical particles.|
|-|-|-|
|![](./assets/cat_image_depth.jpg)|![](./assets/cat_ControlNet_sunshine.jpg)|![](./assets/cat_ControlNet_magic.jpg)|

|Condition|Prompt: A lovely fox wearing a casual green shirt, sitting in a cafe bar, smiling gently, peaceful anime aesthetic.|Prompt: A cute 3D rendered anthropomorphic fox character wearing a bright green shirt, sitting in a cozy magical tavern, smiling happily.|
|-|-|-|
|![](./assets/fox.png)|![](./assets/fox_ControlNet_sunshine.jpg)|![](./assets/fox_ControlNet_magic.jpg)|

|Condition|Prompt: A photorealistic glass crystal ball containing a tiny, dreamy scene of a castle, a large tree, and a girl, soft warm lighting, detailed texture.|Prompt: A cute 3D Pixar style scene inside a crystal ball, featuring a girl standing by a large tree with a castle in the background.|
|-|-|-|
|![](./assets/ball.png)|![](./assets/ball_ControlNet_sunshine.jpg)|![](./assets/ball_ControlNet_magic.jpg)|

## Inference Code

* Install [DiffSynth-Studio](https://github.com/modelscope/DiffSynth-Studio)

```
git clone https://github.com/modelscope/DiffSynth-Studio.git
cd DiffSynth-Studio
pip install -e .
```

* Direct inference (requires 40GB GPU memory)

```python
from diffsynth.diffusion.template import TemplatePipeline
from diffsynth.pipelines.flux2_image import Flux2ImagePipeline, ModelConfig
import torch
from modelscope import dataset_snapshot_download
from PIL import Image
```

```python
pipe = Flux2ImagePipeline.from_pretrained(
    torch_dtype=torch.bfloat16,
    device="cuda",
    model_configs=[
        ModelConfig(model_id="black-forest-labs/FLUX.2-klein-base-4B", origin_file_pattern="transformer/*.safetensors"),
        ModelConfig(model_id="black-forest-labs/FLUX.2-klein-4B", origin_file_pattern="text_encoder/*.safetensors"),
        ModelConfig(model_id="black-forest-labs/FLUX.2-klein-4B", origin_file_pattern="vae/diffusion_pytorch_model.safetensors"),
    ],
    tokenizer_config=ModelConfig(model_id="black-forest-labs/FLUX.2-klein-4B", origin_file_pattern="tokenizer/"),
)
template = TemplatePipeline.from_pretrained(
    torch_dtype=torch.bfloat16,
    device="cuda",
    model_configs=[ModelConfig(model_id="DiffSynth-Studio/Template-KleinBase4B-ControlNet")],
)
dataset_snapshot_download(
    "DiffSynth-Studio/examples_in_diffsynth",
    allow_file_pattern=["templates/*"],
    local_dir="data/examples",
)
image = template(
    pipe,
    prompt="A cat is sitting on a stone, bathed in bright sunshine.",
    seed=0, cfg_scale=4, num_inference_steps=50,
    template_inputs=[{
        "image": Image.open("data/examples/templates/image_depth.jpg"),
        "prompt": "A cat is sitting on a stone, bathed in bright sunshine.",
    }],
    negative_template_inputs=[{
        "image": Image.open("data/examples/templates/image_depth.jpg"),
        "prompt": "",
    }],
)
image.save("image_ControlNet_sunshine.jpg")
image = template(
    pipe,
    prompt="A cat is sitting on a stone, surrounded by colorful magical particles.",
    seed=0, cfg_scale=4, num_inference_steps=50,
    template_inputs=[{
        "image": Image.open("data/examples/templates/image_depth.jpg"),
        "prompt": "A cat is sitting on a stone, surrounded by colorful magical particles.",
    }],
    negative_template_inputs=[{
        "image": Image.open("data/examples/templates/image_depth.jpg"),
        "prompt": "",
    }],
)
image.save("image_ControlNet_magic.jpg")
```

* Enable lazy loading and memory management, requires 24G GPU memory

```python
from diffsynth.diffusion.template import TemplatePipeline
from diffsynth.pipelines.flux2_image import Flux2ImagePipeline, ModelConfig
import torch
from modelscope import dataset_snapshot_download
from PIL import Image
```

```python
vram_config = {
    "offload_dtype": "disk",
    "offload_device": "disk",
    "onload_dtype": torch.float8_e4m3fn,
    "onload_device": "cpu",
    "preparing_dtype": torch.float8_e4m3fn,
    "preparing_device": "cuda",
    "computation_dtype": torch.bfloat16,
    "computation_device": "cuda",
}
pipe = Flux2ImagePipeline.from_pretrained(
    torch_dtype=torch.bfloat16,
    device="cuda",
    model_configs=[
        ModelConfig(model_id="black-forest-labs/FLUX.2-klein-base-4B", origin_file_pattern="transformer/*.safetensors", **vram_config),
        ModelConfig(model_id="black-forest-labs/FLUX.2-klein-4B", origin_file_pattern="text_encoder/*.safetensors", **vram_config),
        ModelConfig(model_id="black-forest-labs/FLUX.2-klein-4B", origin_file_pattern="vae/diffusion_pytorch_model.safetensors"),
    ],
    tokenizer_config=ModelConfig(model_id="black-forest-labs/FLUX.2-klein-4B", origin_file_pattern="tokenizer/"),
    vram_limit=torch.cuda.mem_get_info("cuda")[1] / (1024 ** 3) - 0.5,
)
template = TemplatePipeline.from_pretrained(
    torch_dtype=torch.bfloat16,
    device="cuda",
    model_configs=[ModelConfig(model_id="DiffSynth-Studio/Template-KleinBase4B-ControlNet")],
    lazy_loading=True,
)
dataset_snapshot_download(
    "DiffSynth-Studio/examples_in_diffsynth",
    allow_file_pattern=["templates/*"],
    local_dir="data/examples",
)
image = template(
    pipe,
    prompt="A cat is sitting on a stone, bathed in bright sunshine.",
    seed=0, cfg_scale=4, num_inference_steps=50,
    template_inputs = [{
        "image": Image.open("data/examples/templates/image_depth.jpg"),
        "prompt": "A cat is sitting on a stone, bathed in bright sunshine.",
    }],
    negative_template_inputs = [{
        "image": Image.open("data/examples/templates/image_depth.jpg"),
        "prompt": "",
    }],
)
image.save("image_ControlNet_sunshine.jpg")
image = template(
    pipe,
    prompt="A cat is sitting on a stone, surrounded by colorful magical particles.",
    seed=0, cfg_scale=4, num_inference_steps=50,
    template_inputs = [{
        "image": Image.open("data/examples/templates/image_depth.jpg"),
        "prompt": "A cat is sitting on a stone, surrounded by colorful magical particles.",
    }],
    negative_template_inputs = [{
        "image": Image.open("data/examples/templates/image_depth.jpg"),
        "prompt": "",
    }],
)
image.save("image_ControlNet_magic.jpg")
```

## Training Code

After installing DiffSynth-Studio, use the following script to start training. For more information, please refer to the [DiffSynth-Studio Documentation](https://diffsynth-studio-doc.readthedocs.io/zh-cn/latest/).

```shell
modelscope download --dataset DiffSynth-Studio/diffsynth_example_dataset --include "flux2/Template-KleinBase4B-ControlNet/*" --local_dir ./data/diffsynth_example_dataset

accelerate launch examples/flux2/model_training/train.py \
  --dataset_base_path data/diffsynth_example_dataset/flux2/Template-KleinBase4B-ControlNet \
  --dataset_metadata_path data/diffsynth_example_dataset/flux2/Template-KleinBase4B-ControlNet/metadata.jsonl \
  --extra_inputs "template_inputs" \
  --max_pixels 1048576 \
  --dataset_repeat 50 \
  --model_id_with_origin_paths "black-forest-labs/FLUX.2-klein-4B:text_encoder/*.safetensors,black-forest-labs/FLUX.2-klein-base-4B:transformer/*.safetensors,black-forest-labs/FLUX.2-klein-4B:vae/diffusion_pytorch_model.safetensors" \
  --template_model_id_or_path "DiffSynth-Studio/Template-KleinBase4B-ControlNet:" \
  --tokenizer_path "black-forest-labs/FLUX.2-klein-4B:tokenizer/" \
  --learning_rate 1e-4 \
  --num_epochs 2 \
  --remove_prefix_in_ckpt "pipe.template_model." \
  --output_path "./models/train/Template-KleinBase4B-ControlNet_full" \
  --trainable_models "template_model" \
  --use_gradient_checkpointing \
  --find_unused_parameters
```