model_668764239_mae_giant.py

Model Overview

A giant-scale implementation of the mae architecture, built for retrieval tasks.

Architecture

• Architecture: mae
• Scale: giant
• Attention: dilated
• Fusion strategy: tensor fusion
• Task head: retrieval
• Activation: gelu
• Normalization: instancenorm
• Initialization: xavier

Training

• Optimizer: adafactor
• LR scheduler: exponential

Files

• model_668764239_mae_giant.py — main artifact of this repository

License

See the license field above.