model_072084627_cnn_transformer_giant.py

Model Overview

A giant-scale implementation of the cnn transformer architecture, built for matching tasks.

Architecture

• Architecture: cnn transformer
• Scale: giant
• Attention: sliding window
• Fusion strategy: co attention
• Task head: matching
• Activation: swish
• Normalization: groupnorm
• Initialization: xavier

Training

• Optimizer: lion
• LR scheduler: onecycle

Files

• model_072084627_cnn_transformer_giant.py — main artifact of this repository

License

See the license field above.