Update README.md
Browse files
README.md
CHANGED
|
@@ -25,6 +25,16 @@ So we construct synthesized captions of LAION-aesthetics-V2 6+ by using a large
|
|
| 25 |
KOALA-700M-LLaVA-Caption and KOALA-1B-LLaVA-Caption is trained on the synthesized caption-image pairs of LAION-aesthetics-V2 6+.
|
| 26 |
|
| 27 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 28 |
## Abstract
|
| 29 |
### TL;DR
|
| 30 |
> We propose a fast text-to-image model, called KOALA, by compressing SDXL's U-Net and distilling knowledge from SDXL into our model. KOALA-700M can generate a 1024x1024 image in less than 1.5 seconds on an NVIDIA 4090 GPU, which is more than 2x faster than SDXL. KOALA-700M can be used as a decent alternative between SDM and SDXL in limited resources.
|
|
|
|
| 25 |
KOALA-700M-LLaVA-Caption and KOALA-1B-LLaVA-Caption is trained on the synthesized caption-image pairs of LAION-aesthetics-V2 6+.
|
| 26 |
|
| 27 |
|
| 28 |
+
## KOALA Model Cards
|
| 29 |
+
|
| 30 |
+
|Model|link|
|
| 31 |
+
|:--|:--|
|
| 32 |
+
|koala-700m | https://huggingface.co/etri-vilab/koala-700m|
|
| 33 |
+
|koala-700m-llava-cap | https://huggingface.co/etri-vilab/koala-700m-llava-cap|
|
| 34 |
+
|koala-1b | https://huggingface.co/etri-vilab/koala-1bm|
|
| 35 |
+
|koala-1b-llava-cap | https://huggingface.co/etri-vilab/koala-1b-llava-cap|
|
| 36 |
+
|
| 37 |
+
|
| 38 |
## Abstract
|
| 39 |
### TL;DR
|
| 40 |
> We propose a fast text-to-image model, called KOALA, by compressing SDXL's U-Net and distilling knowledge from SDXL into our model. KOALA-700M can generate a 1024x1024 image in less than 1.5 seconds on an NVIDIA 4090 GPU, which is more than 2x faster than SDXL. KOALA-700M can be used as a decent alternative between SDM and SDXL in limited resources.
|