model_239769637_albef_base.py

Model Overview

A base-scale implementation of the albef architecture, built for retrieval tasks.

Architecture

• Architecture: albef
• Scale: base
• Attention: sliding window
• Fusion strategy: low rank
• Task head: retrieval
• Activation: relu
• Normalization: layernorm
• Initialization: kaiming

Training

• Optimizer: lamb
• LR scheduler: step

Files

• model_239769637_albef_base.py — main artifact of this repository

License

See the license field above.