rlmollm-assets

RLMolLM Pre-trained Assets

This repository contains pre-trained model weights and initial populations for the RLMolLM package.

Contents

Models

• models/moses_pretrained.pt - Pre-trained model on MOSES dataset (417MB)

Initial Populations

• initial_populations/moses_2000.csv - 2000 molecules for initialization (3.6MB)

Usage

These files are automatically downloaded when you use the download_assets.py script from the main repository:

# Clone the main repo
git clone https://github.com/your-org/rlmollm.git
cd rlmollm

# Install
pip install -e .

# Download assets (this repo)
python download_assets.py --dataset moses

Manual Download

If you prefer to download manually:

from huggingface_hub import hf_hub_download

# Download model
model_path = hf_hub_download(
    repo_id="scofieldlinlin/rlmollm-assets",
    filename="models/moses_pretrained.pt"
)

# Download initial population
population_path = hf_hub_download(
    repo_id="scofieldlinlin/rlmollm-assets",
    filename="initial_populations/moses_2000.csv"
)

Citation

If you use these assets, please cite:

@software{rlmollm2024,
  title={RLMolLM: Reinforcement Learning-Enhanced Language Model for Molecular Design},
  author={Your Name},
  year={2024},
  url={https://github.com/your-org/rlmollm}
}

License

Same as the main RLMolLM repository.