Update README.md
Browse files
README.md
CHANGED
|
@@ -5,6 +5,16 @@ pipeline_tag: image-classification
|
|
| 5 |
|
| 6 |
Pytorch weights for Kornia ViT converted from the original google JAX vision-transformer repo.
|
| 7 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 8 |
Original weights from https://github.com/google-research/vision_transformer: This weight is based on the
|
| 9 |
[Original ViT_L/16 pretrained on imagenet21k](https://storage.googleapis.com/vit_models/augreg/L_16-i21k-300ep-lr_0.001-aug_strong1-wd_0.1-do_0.0-sd_0.0.npz)
|
| 10 |
|
|
|
|
| 5 |
|
| 6 |
Pytorch weights for Kornia ViT converted from the original google JAX vision-transformer repo.
|
| 7 |
|
| 8 |
+
Using it with kornia:
|
| 9 |
+
|
| 10 |
+
```python
|
| 11 |
+
from kornia.contrib import VisionTransformer
|
| 12 |
+
|
| 13 |
+
vit_model = VisionTransformer.from_config('vit_l/16', pretrained=True)
|
| 14 |
+
...
|
| 15 |
+
```
|
| 16 |
+
|
| 17 |
+
|
| 18 |
Original weights from https://github.com/google-research/vision_transformer: This weight is based on the
|
| 19 |
[Original ViT_L/16 pretrained on imagenet21k](https://storage.googleapis.com/vit_models/augreg/L_16-i21k-300ep-lr_0.001-aug_strong1-wd_0.1-do_0.0-sd_0.0.npz)
|
| 20 |
|