model_031035109_vit_huge.py

Model Overview

A huge-scale implementation of the vit architecture, built for generation tasks.

Architecture

• Architecture: vit
• Scale: huge
• Attention: multi query
• Fusion strategy: gated fusion
• Task head: generation
• Activation: gelu tanh
• Normalization: batchnorm
• Initialization: kaiming

Training

• Optimizer: lion
• LR scheduler: cosine

Files

• model_031035109_vit_huge.py — main artifact of this repository

License

See the license field above.