project-forecast / README.md
rriveranicholas's picture
Upload 2 files
ea29a00 verified
|
Raw
History Blame Contribute Delete
733 Bytes
metadata
license: apache-2.0
tags:
  - adam
  - blip
  - cross-attention
  - mish
  - onecycle
  - retrieval
  - scalenorm
  - small
  - standard
  - xavier-uniform

pipeline.py

Model Overview

A small-scale implementation of the blip architecture, built for retrieval tasks.

Architecture

  • Architecture: blip
  • Scale: small
  • Attention: standard
  • Fusion strategy: cross attention
  • Task head: retrieval
  • Activation: mish
  • Normalization: scalenorm
  • Initialization: xavier uniform

Training

  • Optimizer: adam
  • LR scheduler: onecycle

Files

  • pipeline.py — main artifact of this repository

License

See the license field above.