model_022077851_vit_large.py

Model Overview

A large-scale implementation of the vit architecture, built for matching tasks.

Architecture

• Architecture: vit
• Scale: large
• Attention: grouped query
• Fusion strategy: cross attention
• Task head: matching
• Activation: relu
• Normalization: scalenorm
• Initialization: xavier

Training

• Optimizer: sgd
• LR scheduler: constant warmup

Files

• model_022077851_vit_large.py — main artifact of this repository

License

See the license field above.