README.md

---
license: other
license_name: flux-non-commercial-license-v2.1
license_link: https://huggingface.co/black-forest-labs/FLUX.2-klein-base-9B/blob/main/LICENSE.md
language:
- en
base_model:
- black-forest-labs/FLUX.2-klein-base-9B
tags:
- text-to-image
- scientific-diagrams
- methodology-figures
- flux
- diffusers
pipeline_tag: text-to-image
---

# SciForma-Base

**SciForma-Base** is the Stage 1+2 SFT checkpoint of [SciForma: Structure-Faithful Generation of Scientific Diagrams](https://github.com/microsoft/SciForma), fine-tuned from [FLUX.2-klein-base-9B](https://huggingface.co/black-forest-labs/FLUX.2-klein-base-9B).

| Model | Overall ↑ | Comp. ↑ | Arrow ↑ | Text ↑ |
|-------|-----------|---------|---------|--------|
| SciForma-9B | 69.51 | 74.49 | 66.46 | 67.00 |
| **SciForma-Base** | **67.59** | 73.52 | 64.64 | 63.84 |
| FLUX.2-klein-base-9B (no fine-tuning) | 33.87 | 42.80 | 25.20 | 23.60 |

Scores on SciFormaBench-2K (GPT-5.4 judge, split_dims=True).

## Usage

This repository contains only the fine-tuned **transformer** weights. The text encoder, tokenizer, VAE, and scheduler are inherited from the base model.

```python
import torch
from diffusers import Flux2KleinPipeline, Flux2Transformer2DModel

# Load fine-tuned transformer
transformer = Flux2Transformer2DModel.from_pretrained(
    "microsoft/SciForma-Base",
    subfolder="transformer",
    torch_dtype=torch.bfloat16,
)

# Load full pipeline with base model components
pipe = Flux2KleinPipeline.from_pretrained(
    "black-forest-labs/FLUX.2-klein-base-9B",
    transformer=transformer,
    torch_dtype=torch.bfloat16,
)
pipe.enable_model_cpu_offload()

# Generate
prompt = "The figure illustrates a transformer architecture with encoder and decoder..."
image = pipe(
    prompt,
    height=576,
    width=1008,
    num_inference_steps=28,
    guidance_scale=4.0,
).images[0]
image.save("output.png")
```

## Training Details

- **Base model**: FLUX.2-klein-base-9B (9.08B parameters)
- **Stage 1**: Full fine-tuning on 661K arXiv methodology diagrams at 768px (200K steps, 8× B200)
- **Stage 2**: Joint generation + editing fine-tuning on 244K high-quality 1024px diagrams + 70K editing triplets (120K steps, 8× B200)
- **Training data**: [microsoft/SciFormaData-700K](https://huggingface.co/datasets/microsoft/SciFormaData-700K)

## License

Released under the [FLUX Non-Commercial License v2.1](https://huggingface.co/black-forest-labs/FLUX.2-klein-base-9B/blob/main/LICENSE.md). Non-commercial research use only.

> This FLUX Model is licensed by Black Forest Labs Inc. under the FLUX Non-Commercial License. Copyright Black Forest Labs Inc. This model has been modified by Microsoft Research Asia.