ShiftML2: A machine learning model of chemical shifts for chemically and structurally diverse molecular solids This record contains the ShiftML2 Python package. Installation First, download all files, unzip the "ShiftML2.zip" archive, and place the model files for the elements you want to predict (.pk files) in the ShiftML2/models directory. A script is available to create a Python environment containing all necessary packages to run ShiftML2. To install it, run the file: ./install_env.sh If you already have a python environment containing the required packages, you can directly install the ShiftML2 package using: pip install . You can then test the installation by running the "example.ipynb" notebook. Additional files We also provide the complete initial and final (after outlier removal and FPS selection) training set, as well as the test set with ShiftML2 predictions in the files "initial_training_test_set.zip" and "final_training_test_set.zip". The training scripts are provided as jupyter notebooks in the file "training_scripts.zip". The relaxed structures for the experimental benchmark set are in the "Experimental_benchmark_nmr.zip" file. The structures of the sets of candidate polymorphs are in the "ShiftML_poly.zip" file. The relaxed and MD structures for comparing ShiftML2 accuracies between relaxed and MD structures of AZD5718, AZD8321 and cocaine are in the "relax_vs_md.zip" file.