README.md

metadata

library_name: diffusers

Flux2 Klein 4b Modular repo

This contains mellon config for A single-node Mellon block for the FLUX.2-klein-4B pipeline.

How to create a modular repo like this one?

load the pipeline and save it to a new Hub repository:

from diffusers import ModularPipeline

pipe = ModularPipeline.from_pretrained("black-forest-labs/FLUX.2-klein-4B")
pipe.save_pretrained(
    "local/dir",
    repo_id="your-username/flux2-klein-auto-blocks",
    push_to_hub=True
)

Generate the Mellon Config

Create and push the Mellon configuration with your desired UI types:

from diffusers.modular_pipelines.mellon_node_utils import MellonPipelineConfig

mellon_config = MellonPipelineConfig.from_custom_block(
    pipe.blocks,
    input_types={
        "prompt": "textbox",
        "image": "image",
        "num_inference_steps": "number",
    },
    output_types={
        "images": "image"
    }
)

mellon_config.save(
    "local/dir",
    repo_id="your-username/flux2-klein-auto-blocks",
    push_to_hub=True
)

Review and Adjust

The generated mellon_pipeline_config.json may need adjustments. You can:

• Remove inputs you don't want exposed in the UI
• Add slider ranges for numeric inputs
• Add dropdown options for string inputs
• update default value for numeric inputs

Use in Mellon

1. Drag a Dynamic Block Node from the ModularDiffusers section
2. Enter diffusers/FLUX.2-klein-4B-modular and click "LOAD CUSTOM BLOCK"
3. The node will show your configured inputs (prompt, image, num_inference_steps) and outputs (images)

Supported Mellon Types

Type	Input/Output	Description
image	Both	Image (PIL Image)
textbox	Input	Text input
number	Input	Numeric input
slider	Input	Slider for numeric values
dropdown	Input	Dropdown selection
checkbox	Input	Boolean toggle
text	Both	Text display