efficientvit

Efficient vision foundation models for high-resolution generation and perception.

This is a warehouse for EfficientViT-pytorch-model, can be used to train your dataset

End-to-end MLOps pipeline for multimodal e-commerce product classification (text + image) — ingestion, training, inference and monitoring.

Pretraining the EfficientViT-B4 model on the ImageNet-1k dataset

KTPFormer(CVPR2024)에 EfficientViT의 Cascaded Group Attention을 이식해 GFLOPs 29.5% 절감 — Q/K 차원 축소와 Nonlinear Feature Augmentation 실험 기록