Add gMLP-S weights, 79.6 top-1

vit_and_bit_test_fixes
Ross Wightman 3 years ago
parent 85f894e03d
commit 20a2be14c3

@ -23,6 +23,9 @@ I'm fortunate to be able to dedicate significant time and money of my own suppor
## What's New
### June 23, 2021
* Reproduce gMLP model training, `gmlp_s16_224` trained to 79.6 top-1, matching [paper](https://arxiv.org/abs/2105.08050).
### June 20, 2021
* Release Vision Transformer 'AugReg' weights from [How to train your ViT? Data, Augmentation, and Regularization in Vision Transformers](https://arxiv.org/abs/2106.10270)
* .npz weight loading support added, can load any of the 50K+ weights from the [AugReg series](https://console.cloud.google.com/storage/browser/vit_models/augreg)

@ -129,7 +129,9 @@ default_cfgs = dict(
mean=IMAGENET_DEFAULT_MEAN, std=IMAGENET_DEFAULT_STD),
gmlp_ti16_224=_cfg(),
gmlp_s16_224=_cfg(),
gmlp_s16_224=_cfg(
url='https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/gmlp_s16_224_raa-10536d42.pth',
),
gmlp_b16_224=_cfg(),
)

Loading…
Cancel
Save