Clone the source code from GitHub:
git clone https://github.com/JeixHuang/DEJO.git
cd DEJONative Runner: Setup a conda environment using conda / mamba:
conda conda env create --file conda-recipe.yaml # or `mamba env create --file conda-recipe.yaml`This will automatically setup all dependencies.
Containerized Runner: Other than using the native machine with conda isolation, as an alternative, you can also use docker images to configure the environment.
Firstly, please follow NVIDIA Container Toolkit: Installation Guide and NVIDIA Docker: Installation Guide to setup nvidia-docker.
Then you can run:s
make docker-runThis command will build and start a docker container installed with proper dependencies.
The host path / will be mapped to /host and the current working directory will be mapped to /workspace inside the container.
-
Install required packages:
-
Load or generate embeddings and save them in
data/embeddings.npy. -
Run the main program:
data/load_data.py: Functions to load and save embeddings.models/linear_transform.py: Linear transformation model.models/gan.py: GAN model.utils/evaluation.py: Evaluation functions for similarity and harmfulness.main.py: Main script to run the entire process.
- Cosine similarity is used to evaluate the semantic similarity between generated and target embeddings.
- A placeholder harmfulness model is used to evaluate the harmfulness of generated embeddings.
- Improve the harmfulness evaluation model.
- Experiment with different architectures and hyperparameters for GAN.