mirror of
https://github.com/huggingface/candle.git
synced 2025-06-15 18:28:24 +00:00
29e25c458d
* correct optional SE layer dimensions. * head_dim instead of num_heads is 32. * update test example output.
candle-fastvit
FastViT: A Fast Hybrid Vision Transformer using Structural Reparameterization. This candle implementation uses a pre-trained FastViT network for inference. The classification head has been trained on the ImageNet dataset and returns the probabilities for the top-5 classes.
Running an example
$ cargo run --example fastvit --release -- --image candle-examples/examples/yolo-v8/assets/bike.jpg --which sa12
loaded image Tensor[dims 3, 256, 256; f32]
model built
mountain bike, all-terrain bike, off-roader: 52.67%
bicycle-built-for-two, tandem bicycle, tandem: 7.93%
unicycle, monocycle : 3.46%
maillot : 1.32%
crash helmet : 1.28%