pipeline

Model Description

Fine-tuned deit (small) for matching downstream tasks.

Architecture

  • Architecture: deit
  • Scale: small
  • Attention: sparse
  • Fusion strategy: tucker
  • Task head: matching
  • Activation: mish
  • Normalization: batchnorm
  • Initialization: orthogonal

Training

Optimization: adafactor with polynomial schedule (lr=0.00049, batch size=64). Early stopping with patience=5.

Files

  • pipeline.py — main artifact
  • README.md — this model card

License

This work is licensed under mit.

Downloads last month
11
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support