qlora-detector / README.md
leonardosouza's picture
Upload 2 files
8b39132 verified
|
Raw History Blame Contribute Delete
729 Bytes
metadata
license: cc-by-4.0
tags:
  - concat-mlp
  - contrastive
  - dilated
  - gelu
  - groupnorm
  - lamb
  - perceiver
  - polynomial
  - small
  - xavier

pipeline.py

Model Overview

A small-scale implementation of the perceiver architecture, built for contrastive tasks.

Architecture

  • Architecture: perceiver
  • Scale: small
  • Attention: dilated
  • Fusion strategy: concat mlp
  • Task head: contrastive
  • Activation: gelu
  • Normalization: groupnorm
  • Initialization: xavier

Training

  • Optimizer: lamb
  • LR scheduler: polynomial

Files

  • pipeline.py — main artifact of this repository

License

See the license field above.