A CKPT file contains a dump of a PyTorch Lightning machine learning model. It preserves the states of a model during or after training.
Tensorflow also uses .ckpt and .pb files to save models. To save models in TensorFlow, use saver.save(sess, “final-model.ckpt”). This saves session variables to files. TensorFlow checkpoint files store model variable values. To load TensorFlow checkpoints, use CKPT file extensions like *.ckpt.meta, *.ckpt.index, *.ckpt.data.
.ckpt files load models slowly. To improve security, online apps are available that scan pickle files to ensure .ckpt files are clean.