Skip to content

Latest commit

 

History

History
16 lines (11 loc) · 937 Bytes

File metadata and controls

16 lines (11 loc) · 937 Bytes

Neuro-Visualizer: An Auto-encoder-based Loss Landscape Visualization Method

Important links:

This link contains both the necessary data for reproducibility and the results of the experiemnts conducted in our paper.

  • Input trajectory model data: trajectories_v3.zip
  • Trained Neuro-Visualizers and results: saved_models_v3.zip
  • Other important auxiliary data: data_v3.zip

How to run an experiment

  • git clone this repo.
  • pip install requirements.txt
  • Extract the zip files above inside your repository.
  • cd scripts and then bash scripts\*.sh for the experiment of interest.