TorchCFM is a conditional flow matching library built on PyTorch and Lightning. It provides implementations of flow matching techniques that use optimal transport theory to construct continuous paths between probability distributions for generative modeling. The library supports training and evaluation workflows for flow-based models.