metadata
license: cc-by-4.0
tags:
- bilinear
- clip
- contrastive
- exponential
- kaiming-normal
- lion
- mish
- rmsnorm
- small
- sparse
model_196640384_clip_small.py
Model Overview
A small-scale implementation of the clip architecture, built for contrastive tasks.
Architecture
- Architecture: clip
- Scale: small
- Attention: sparse
- Fusion strategy: bilinear
- Task head: contrastive
- Activation: mish
- Normalization: rmsnorm
- Initialization: kaiming normal
Training
- Optimizer: lion
- LR scheduler: exponential
Files
model_196640384_clip_small.py— main artifact of this repository
License
See the license field above.