model_488174279_mixer_tiny.py

Model Overview

A tiny-scale implementation of the mixer architecture, built for contrastive tasks.

Architecture

• Architecture: mixer
• Scale: tiny
• Attention: standard
• Fusion strategy: tensor fusion
• Task head: contrastive
• Activation: relu
• Normalization: groupnorm
• Initialization: kaiming normal

Training

• Optimizer: novograd
• LR scheduler: constant warmup

Files

• model_488174279_mixer_tiny.py — main artifact of this repository

License

See the license field above.