Update README.md
Browse files
README.md
CHANGED
|
@@ -107,3 +107,14 @@ gen_images = pipe(
|
|
| 107 |
gen_images.save('out_lin.png')
|
| 108 |
```
|
| 109 |
<a href="https://huggingface.co/Adapter/t2iadapter/resolve/main/figs_SDXLV1.0/cond_lin.png"><img width="480" style="margin:0;padding:0;" src="https://huggingface.co/Adapter/t2iadapter/resolve/main/figs_SDXLV1.0/res_lin.png"/></a>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 107 |
gen_images.save('out_lin.png')
|
| 108 |
```
|
| 109 |
<a href="https://huggingface.co/Adapter/t2iadapter/resolve/main/figs_SDXLV1.0/cond_lin.png"><img width="480" style="margin:0;padding:0;" src="https://huggingface.co/Adapter/t2iadapter/resolve/main/figs_SDXLV1.0/res_lin.png"/></a>
|
| 110 |
+
|
| 111 |
+
### Training
|
| 112 |
+
|
| 113 |
+
Our training script was built on top of the official training script that we provide [here](https://github.com/huggingface/diffusers/blob/main/examples/t2i_adapter/README_sdxl.md).
|
| 114 |
+
|
| 115 |
+
The model is trained on 3M high-resolution image-text pairs from LAION-Aesthetics V2 with
|
| 116 |
+
|
| 117 |
+
- Training steps: 20000
|
| 118 |
+
- Batch size: Data parallel with a single gpu batch size of `16` for a total batch size of `256`.
|
| 119 |
+
- Learning rate: Constant learning rate of `1e-5`.
|
| 120 |
+
- Mixed precision: fp16
|