CvT-W24

Wu et al., 2021 — CvT: Introducing Convolutions to Vision Transformers (arXiv:2103.15808)

Lucid port of transformers/microsoft/cvt-w24-384-22k, converted to Lucid-native safetensors.

Available weights

Tag	acc@1	acc@5	Params	GFLOPs	Size	Source
IN22K_FT_IN1K_384 (default)	87.7	—	277.2M	—	1058.02 MB	transformers

Usage

import lucid.models as models
from lucid.models.weights import CvTW24Weights

# default tag
model = models.cvt_w24_cls(pretrained=True)

# explicit tag (enum or string)
model = models.cvt_w24_cls(weights=CvTW24Weights.IN22K_FT_IN1K_384)
model = models.cvt_w24_cls(pretrained="IN22K_FT_IN1K_384")

# preprocessing travels with the weights
weights = CvTW24Weights.IN22K_FT_IN1K_384
preprocess = weights.transforms()
logits = model(preprocess(image)[None]).logits

Conversion

Converted from transformers/microsoft/cvt-w24-384-22k via python -m tools.convert_weights cvt_w24 --tag IN22K_FT_IN1K_384. Key mapping + numerical parity verified against the source.

License

apache-2.0 — inherited from the original weights.

Citation

@inproceedings{wu2021cvt,
  title={CvT: Introducing Convolutions to Vision Transformers},
  author={Wu, Haiping and Xiao, Bin and Codella, Noel and Liu, Mengchen and Dai, Xiyang and Yuan, Lu and Zhang, Lei},
  booktitle={ICCV}, year={2021}
}