metadata
license: apache-2.0
tags:
- approx-gelu
- clip
- concat-mlp
- flash
- groupnorm
- kaiming-normal
- lamb
- large
- matching
- polynomial
model_206285055_clip_large.py
Model Overview
A large-scale implementation of the clip architecture, built for matching tasks.
Architecture
- Architecture: clip
- Scale: large
- Attention: flash
- Fusion strategy: concat mlp
- Task head: matching
- Activation: approx gelu
- Normalization: groupnorm
- Initialization: kaiming normal
Training
- Optimizer: lamb
- LR scheduler: polynomial
Files
model_206285055_clip_large.py— main artifact of this repository
License
See the license field above.