README.md

---
license: mit
library_name: sd15-flow-trainer
tags:
  - geometric-deep-learning
  - stable-diffusion
  - ksimplex
  - pentachoron
  - flow-matching
  - cross-attention-prior
base_model: sd-legacy/stable-diffusion-v1-5
pipeline_tag: text-to-image
---

# V1 weights test push

https://github.com/AbstractEyes/sd15-flow-trainer

https://huggingface.co/AbstractPhil/sd15-rectified-geometric-matching/blob/main/colab_trainer.py

Step 0

![image](https://cdn-uploads.huggingface.co/production/uploads/630cf55b15433862cfc9556f/L8W0i4EzWc2XKKU3YImzp.png)

Step 500

![image](https://cdn-uploads.huggingface.co/production/uploads/630cf55b15433862cfc9556f/cexS-dFaojxUebsW1KqR4.png)


Step 1000

![image](https://cdn-uploads.huggingface.co/production/uploads/630cf55b15433862cfc9556f/wFZsCCiTawdTMQZKtEs9x.png)

Step 1500

![image](https://cdn-uploads.huggingface.co/production/uploads/630cf55b15433862cfc9556f/9polT0bIUJ5ypM4PJLsOk.png)


# KSimplex Geometric Attention Prior

Geometric cross-attention prior for SD1.5 using pentachoron (4-simplex) structures.

## Architecture

| Component | Params |
|-----------|--------|
| SD1.5 UNet (frozen) | 859,520,964 |
| **Geo prior (trained)** | **4,845,725** |

The geometric prior modulates CLIP encoder hidden states through
4-layer stacked k-simplex attention before they reach
the 16 cross-attention blocks in the UNet.

## Simplex Configuration

| Parameter | Value |
|-----------|-------|
| k (simplex dim) | 4 |
| Embedding dim | 32 |
| Feature dim | 768 |
| Stacked layers | 4 |
| Attention heads | 8 |
| Base deformation | 0.25 |
| Residual blend | learnable |
| Timestep conditioned | True |

## Usage

```python
from sd15_trainer_geo.pipeline import load_pipeline, load_geo_from_hub

# Load base SD1.5 + fresh geo prior
pipe = load_pipeline()

# Load trained geo weights from this repo
load_geo_from_hub(pipe, "AbstractPhil/sd15-rectified-geometric-matching")

# Or one-shot: load base + geo in one call
pipe = load_pipeline(geo_repo_id="AbstractPhil/sd15-rectified-geometric-matching")
```

## Training Info

- **dataset**: AbstractPhil/imagenet-synthetic (flux_schnell_512)
- **samples**: 10000
- **epochs**: 1
- **shift**: 2.5
- **base_lr**: 0.0001
- **min_snr_gamma**: 5.0
- **cfg_dropout**: 0.1
- **batch_size**: 6
- **loss_final**: 0.3784324672818184

## Post Analysis


![image](https://cdn-uploads.huggingface.co/production/uploads/630cf55b15433862cfc9556f/qs2vvoY7f9HdfYYuGI-k5.png)

![image](https://cdn-uploads.huggingface.co/production/uploads/630cf55b15433862cfc9556f/5MpURgWYrFmxpZf8KPQWG.png)

![image](https://cdn-uploads.huggingface.co/production/uploads/630cf55b15433862cfc9556f/ofgomH4SkBbyAtcQDeLBn.png)

![image](https://cdn-uploads.huggingface.co/production/uploads/630cf55b15433862cfc9556f/hxA8Q0rsm6wYpQDgB4puQ.png)



## License

MIT — [AbstractPhil](https://huggingface.co/AbstractPhil)