Ambarella/EfficientViT

[image]

EfficientViT is a new family of high-resolution vision models with novel multi-scale linear attention. As such, EfficientViT delivers remarkable performance gains over previous state-of-the-art models with significant speedup on diverse hardware platforms, including mobile CPU, edge GPU, and cloud GPU.

Original paper: EfficientViT: Multi-Scale Linear Attention for High-Resolution Dense Prediction

EfficientViT-L2

EfficientViT is a new family of vision models for efficient high-resolution dense prediction. The core building block of EfficientViT is a new lightweight multi-scale linear attention module that achieves global receptive field and multi-scale learning with only hardware-efficient operations.

Model Configuration:

• —Reference implementation: EfficientViT-L2
• —Original Weight: l2.pt
• —Resolution: 3x512x512
• —Support Cooper version:
• —Cooper SDK: [2.5.4]
• —Cooper Foundry: [2.3]