metadata
license: bsd-3-clause
tags:
- cosine
- gelu-tanh
- instancenorm
- lamb
- large
- linear
- perceiver
- retrieval
- tucker
- xavier
eval.py
Model Overview
A large-scale implementation of the perceiver architecture, built for retrieval tasks.
Architecture
- Architecture: perceiver
- Scale: large
- Attention: linear
- Fusion strategy: tucker
- Task head: retrieval
- Activation: gelu tanh
- Normalization: instancenorm
- Initialization: xavier
Training
- Optimizer: lamb
- LR scheduler: cosine
Files
eval.py— main artifact of this repository
License
See the license field above.