File size: 395 Bytes
15705a4
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
# EMMeT

**E**mbedding **M**ixup for **Me**ta-**T**raining (EMMeT), a novel regularization technique that creates new episodic batches through an embedding shuffling and averaging regime.

Weights for the CONV4 and DINO-S models are provided for EMMeT evaluation.

For usage and instructions, please refer to the Github repository: https://github.com/Brikwerk/EMMeT

## Model Date

February 2023