model_662575389_mae_huge.py

Model Overview

A huge-scale implementation of the mae architecture, built for contrastive tasks.

Architecture

• Architecture: mae
• Scale: huge
• Attention: multi query
• Fusion strategy: co attention
• Task head: contrastive
• Activation: swish
• Normalization: groupnorm
• Initialization: kaiming

Training

• Optimizer: lion
• LR scheduler: cosine

Files

• model_662575389_mae_huge.py — main artifact of this repository

License

See the license field above.