MBZUAI / swiftformer-l3

SwiftFormer (swiftformer-l3)

Model description

The SwiftFormer model was proposed in SwiftFormer: Efficient Additive Attention for Transformer-based Real-time Mobile Vision Applications by Abdelrahman Shaker, Muhammad Maaz, Hanoona Rasheed, Salman Khan, Ming-Hsuan Yang, Fahad Shahbaz Khan.

SwiftFormer paper introduces a novel efficient additive attention mechanism that effectively replaces the quadratic matrix multiplication operations in the self-attention computation with linear element-wise multiplications. A series of models called 'SwiftFormer' is built based on this, which achieves state-of-the-art performance in terms of both accuracy and mobile inference speed. Even their small variant achieves 78.5% top-1 ImageNet1K accuracy with only 0.8 ms latency on iPhone 14, which is more accurate and 2× faster compared to MobileViT-v2.

Intended uses & limitations

How to use

import requests
from PIL import Image

url = 'http://images.cocodataset.org/val2017/000000039769.jpg'
image = Image.open(requests.get(url, stream=True).raw)

from transformers import ViTImageProcessor
processor = ViTImageProcessor.from_pretrained('shehan97/swiftformer-l3')
inputs = processor(images=image, return_tensors="pt")


from transformers.models.swiftformer import SwiftFormerForImageClassification
new_model = SwiftFormerForImageClassification.from_pretrained('shehan97/swiftformer-l3')

output = new_model(inputs['pixel_values'], output_hidden_states=True)
logits = output.logits
predicted_class_idx = logits.argmax(-1).item()
print("Predicted class:", new_model.config.id2label[predicted_class_idx])

Limitations and bias

Training data

The classification model is trained on the ImageNet-1K dataset.

Training procedure

Evaluation results