FlowSampler-GPT: Accelerate Rare-Event Molecular Sampling
Your statistical mechanics assistant for designing and tuning normalizing flow models to enhance sampling in molecular simulations, inspired by Grant Rotskoff.
Access FlowSampler-GPTWhat is FlowSampler-GPT?
FlowSampler-GPT is a statistical mechanics assistant focused on designing and tuning normalizing flow models (e.g., Boltzmann Generators) to accelerate the sampling of rare events in molecular simulations. It aids computational chemists and physicists in model architecture choices, training data incorporation, interpreting free-energy landscapes, and applying learned biases in molecular dynamics.
Enhance Your Molecular Simulations with Normalizing Flows
Expert Model Design & Training
Explains choices in normalizing flow architecture (latent space dimensionality, coupling layers) and guides incorporation of trajectory data (e.g., from OpenMM) for training.
Interpret Free Energy & Bias
Helps interpret free-energy convergence plots and advises if more training or a different biasing approach is needed to accurately capture the system's landscape.
Automated Model Generation
Generates PyTorch Lightning templates for normalizing flows (e.g., Boltzmann Generators) and performs hyperparameter grid searches for effective models.
Output Biasing for MD
Computes biasing forces from trained normalizing flow models and outputs PLUMED input files to apply these forces in molecular dynamics simulations.
Clear, Formatted Outputs
Provides training summaries, JSON of best hyperparameters, PLUMED input snippets, and free-energy profiles (CSV/table), all clearly labeled.
Secure & Compliant Operation
Removes proprietary structures (e.g., ligand coordinates) after use and cautions on export-controlled parameters or unpublished data sharing.
Ready to Supercharge Your Molecular Sampling?
Discover how FlowSampler-GPT can help you design efficient normalizing flow models to explore complex molecular landscapes and accelerate the discovery of rare events.
Explore Advanced Sampling