---
license: openrail
datasets:
- laion/laion2B-en-aesthetic
language:
- en
---

Based on https://github.com/lllyasviel/ControlNet/discussions/318

```
accelerate launch train_controlnet.py ^
  --pretrained_model_name_or_path="runwayml/stable-diffusion-v1-5" ^
  --output_dir="control-edgedrawing-default-drop50-fp16/" ^
  --dataset_name="mydataset" ^
  --mixed_precision="fp16" ^
  --proportion_empty_prompts=0.5 ^
  --resolution=512 ^
  --learning_rate=1e-5 ^
  --train_batch_size=1 ^
  --gradient_accumulation_steps=4 ^
  --gradient_checkpointing ^
  --use_8bit_adam ^
  --enable_xformers_memory_efficient_attention ^
  --set_grads_to_none ^
  --seed=0
```

Trained for 40000 steps on images converted with https://github.com/shaojunluo/EDLinePython using `smoothed = False` and default settings:

```
{ 'ksize'            :  5
, 'sigma'            :  1.0
, 'gradientThreshold': 36
, 'anchorThreshold'  :  8
, 'scanIntervals'    :  1
}
```

**TODO**

Results are not good so far:

* `--proportion_empty_prompts=0.5` me be too excessive for 40000 steps
* Use `smoothed = True` next time, maybe control net doesn't pick up on single pixels
* Find better parameter spread instead of default values, most images are very sparse
* Train on more steps
* Train on more diverse dataset
* Train on higher-precision