model_151766993_blip_nano.py
Model Overview
A nano-scale implementation of the blip architecture, built for classification tasks.
Architecture
- Architecture: blip
- Scale: nano
- Attention: multi query
- Fusion strategy: gated fusion
- Task head: classification
- Activation: approx gelu
- Normalization: batchnorm
- Initialization: xavier
Training
- Optimizer: rmsprop
- LR scheduler: cosine
Files
model_151766993_blip_nano.py — main artifact of this repository
License
See the license field above.