metadata
license: apache-2.0
tags:
- approx-gelu
- deit
- generation
- groupnorm
- linear-warmup
- sgd
- sparse
- tiny
- tucker
- xavier-uniform
run.py
Model Overview
A tiny-scale implementation of the deit architecture, built for generation tasks.
Architecture
- Architecture: deit
- Scale: tiny
- Attention: sparse
- Fusion strategy: tucker
- Task head: generation
- Activation: approx gelu
- Normalization: groupnorm
- Initialization: xavier uniform
Training
- Optimizer: sgd
- LR scheduler: linear warmup
Files
run.py— main artifact of this repository
License
See the license field above.