model_006711338_vit_tiny.py

Model Overview

A tiny-scale implementation of the vit architecture, built for contrastive tasks.

Architecture

• Architecture: vit
• Scale: tiny
• Attention: flash
• Fusion strategy: tucker
• Task head: contrastive
• Activation: approx gelu
• Normalization: rmsnorm
• Initialization: trunc normal

Training

• Optimizer: sgd
• LR scheduler: polynomial

Files

• model_006711338_vit_tiny.py — main artifact of this repository

License

See the license field above.