/ Machine Learning

A simple way to achieve SOTA in vision classification with only a single transformer encoder

A simple way to achieve SOTA in vision classification with only a single transformer encoder

Vision Transformer - Pytorch

Implementation of Vision Transformer, a simple way to achieve SOTA in vision classification with only a single transformer encoder, in Pytorch. Significance is further explained in Yannic Kilcher's video. There's really not much to code here, but may as well lay it out for everyone so we expedite the attention revolution.

Install

$ pip install vit-pytorch

Usage

import torch
from vit_pytorch import ViT

v = ViT(
    image_size = 256,
    patch_size = 32,
    num_classes = 1000,
    dim = 1024,
    depth = 6,
    heads = 8,
    mlp_dim = 2048
)

img = torch.randn(1, 3, 256, 256)
preds = v(img) # (1, 1000)

Suggestion

You can train this with a near SOTA self-supervised learning technique, BYOL, with the following code.

(1)

$ pip install byol-pytorch

(2)

import torch
from vit_pytorch import ViT
from byol_pytorch import BYOL

model = ViT(
    image_size = 256,
    patch_size = 32,
    num_classes = 1000,
    dim = 1024,
    depth = 6,
    heads = 8,
    mlp_dim = 2048
)

learner = BYOL(
    model,
    image_size = 256,
    hidden_layer = 'to_cls_token'
)

opt = torch.optim.Adam(learner.parameters(), lr=3e-4)

def sample_unlabelled_images():
    return torch.randn(20, 3, 256, 256)

for _ in range(100):
    images = sample_unlabelled_images()
    loss = learner(images)
    opt.zero_grad()
    loss.backward()
    opt.step()
    learner.update_moving_average() # update moving average of target encoder

# save your improved network
torch.save(model.state_dict(), './pretrained-net.pt')

A pytorch-lightning script is ready for you to use at the repository link above.

Citations

@inproceedings{
    anonymous2021an,
    title={An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale},
    author={Anonymous},
    booktitle={Submitted to International Conference on Learning Representations},
    year={2021},
    url={https://openreview.net/forum?id=YicbFdNTTy},
    note={under review}
}

GitHub

Comments