model_188356023_clip_nano.py
Model Overview
A nano-scale implementation of the clip architecture, built for matching tasks.
Architecture
- Architecture: clip
- Scale: nano
- Attention: sliding window
- Fusion strategy: gated fusion
- Task head: matching
- Activation: gelu tanh
- Normalization: batchnorm
- Initialization: orthogonal
Training
- Optimizer: lion
- LR scheduler: step
Files
model_188356023_clip_nano.py — main artifact of this repository
License
See the license field above.
