metadata
license: bsd-3-clause
tags:
- approx-gelu
- coca
- concat-mlp
- dilated
- generation
- linear-warmup
- orthogonal
- rmsprop
- scalenorm
- small
model_273659226_coca_small.py
Model Overview
A small-scale implementation of the coca architecture, built for generation tasks.
Architecture
- Architecture: coca
- Scale: small
- Attention: dilated
- Fusion strategy: concat mlp
- Task head: generation
- Activation: approx gelu
- Normalization: scalenorm
- Initialization: orthogonal
Training
- Optimizer: rmsprop
- LR scheduler: linear warmup
Files
model_273659226_coca_small.py— main artifact of this repository
License
See the license field above.