/embedding-scope

Interpret and control dense embedding via sparse autoencoder.

Primary LanguagePythonMIT LicenseMIT

Embedding Scope

Table of Contents

Getting Started

  1. Install Conda:

    If Conda isn't installed, download it here.

    Conda is required to manage dependencies and ensure a consistent environment.

  2. Create Environment:

    conda env create -f environment.yml

  3. Activate Environment:

    conda activate scope

  4. Configure Workspace:

    To configure the workspace where the datasets and trained checkpoints are stored, update the source/__init__.py file with the desired path.

    For example:

    from pathlib import Path

workspace = Path("/your/path/to/workspace")
workspace.mkdir(mode=0o770, parents=True, exist_ok=True)

You're all set to begin!