Codes for paper: DrugPlayGround: Pioneering the Benchmark of Large Language Models and Embeddings for Drug Discovery and Applications
To query LLM descriptions, you must have api access for these models:
GPT-4o (OpenAI models, etc.), Claude, Gemini, Mistral, Gemma, and Qwen3. Gemma and Qwen3 are based on Huggingface.
To train classifer or predictor, please refer torch lightning, and install torch lightning.
To train perturbation predictor, plrease refer ChemCPA.
Please refer the codes under different folders for running the benchmarking codes.