model_594432219_deit_giant.py

Model Overview

A giant-scale implementation of the deit architecture, built for contrastive tasks.

Architecture

• Architecture: deit
• Scale: giant
• Attention: dilated
• Fusion strategy: tensor fusion
• Task head: contrastive
• Activation: approx gelu
• Normalization: groupnorm
• Initialization: xavier

Training

• Optimizer: adamw
• LR scheduler: exponential

Files

• model_594432219_deit_giant.py — main artifact of this repository

License

See the license field above.