File size: 1,829 Bytes
ecbb796 fb1ad26 e3431df fb1ad26 ecbb796 fb1ad26 e3431df fb1ad26 e3431df | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 | ---
license: creativeml-openrail-m
library_name: diffusers
tags:
- stable-diffusion
- sd-1.5
- hyper-sd
- onnx
- text-to-image
- 4-step
base_model: emilianJR/epiCRealism
pipeline_tag: text-to-image
---
# epiCRealism + Hyper-SD (4-step) — ONNX
ONNX export of [emilianJR/epiCRealism](https://huggingface.co/emilianJR/epiCRealism) with the [ByteDance/Hyper-SD](https://huggingface.co/ByteDance/Hyper-SD) 4-step LoRA fused into the UNet. SD 1.5 architecture, 512×512 native, Euler scheduler, CFG = 1, **4 steps**.
epiCRealism leans heavily toward photorealistic environments and natural lighting. Pick this one over Realistic Vision when the subject is "a place" rather than "a person."
Converted artifact, not a new model. Training credit: emilianJR (epiCRealism), ByteDance (Hyper-SD).
## What this repo contains
```
model_index.json
feature_extractor/
scheduler/
text_encoder/
tokenizer/
unet/ # epiCRealism UNet + Hyper-SD-15 4-step LoRA fused in
vae_decoder/
vae_encoder/
```
## How it was produced
1. Load `emilianJR/epiCRealism` via `diffusers`.
2. Load `ByteDance/Hyper-SD/Hyper-SD15-4steps-lora.safetensors` via `peft`, `fuse_lora()` into UNet.
3. Save fused pipeline; run `optimum-cli export onnx`.
Toolchain: `optimum 1.24.0`, `diffusers 0.31.0`, `transformers 4.45.2`, `torch 2.4.x` (CUDA 12.4). Conversion script: [`scripts/export-epicrealism-hyper.ps1`](https://github.com/HeliosophLLC/DatumIngest/blob/main/scripts/export-epicrealism-hyper.ps1).
## Inference notes
| Setting | Value |
|---|---|
| Scheduler | Euler |
| Steps | 4 |
| CFG / guidance scale | 1.0 |
| Negative prompt | Skip |
| Resolution | 512×512 native |
## License
CreativeML OpenRAIL-M, inherited from SD 1.5 + epiCRealism + Hyper-SD. License files included. By using this model you accept those terms.
|