EfficientFormer-L3

Li et al., 2022 — EfficientFormer: Vision Transformers at MobileNet Speed (arXiv:2206.01191)

Lucid port of timm/efficientformer_l3.snap_dist_in1k, converted to Lucid-native safetensors.

Available weights

Tag	acc@1	acc@5	Params	GFLOPs	Size	Source
SNAP_DIST_IN1K (default)	82.4	—	31.4M	—	120.07 MB	timm

Usage

import lucid.models as models
from lucid.models.weights import EfficientFormerL3Weights

# default tag
model = models.efficientformer_l3_cls(pretrained=True)

# explicit tag (enum or string)
model = models.efficientformer_l3_cls(weights=EfficientFormerL3Weights.SNAP_DIST_IN1K)
model = models.efficientformer_l3_cls(pretrained="SNAP_DIST_IN1K")

# preprocessing travels with the weights
weights = EfficientFormerL3Weights.SNAP_DIST_IN1K
preprocess = weights.transforms()
logits = model(preprocess(image)[None]).logits

Conversion

Converted from timm/efficientformer_l3.snap_dist_in1k via python -m tools.convert_weights efficientformer_l3 --tag SNAP_DIST_IN1K. Key mapping + numerical parity verified against the source.

License

apache-2.0 — inherited from the original weights.

Citation

@article{li2022efficientformer,
  title={EfficientFormer: Vision Transformers at MobileNet Speed},
  author={Li, Yanyu and Yuan, Geng and Wen, Yang and Hu, Eric and Evangelidis, Georgios and Tulyakov, Sergey and Wang, Yanzhi and Ren, Jian},
  journal={Advances in Neural Information Processing Systems},
  volume={35}, year={2022}
}