The Lightning ModelCheckpoint has an argument save_top_k

save_top_k (int) – if save_top_k == k, the best k models according to the quantity monitored will be saved. If save_top_k == 0, no models are saved. If save_top_k == -1, all models are saved. Please note that the monitors are checked every every_n_epochs epochs. If save_top_k >= 2 and the callback is called multiple times inside an epoch, and the filename remains unchanged, the name of the saved file will be appended with a version count starting with v1 to avoid collisions unless enable_version_counter is set to False. The version counter is unrelated to the top-k ranking of the checkpoint, and we recommend formatting the filename to include the monitored metric to avoid collisions.

METL training sets save_top_k=-1 here

We do not expect to have versioned "last" checkpoints, but they were observed under certain conditions while training in AWS. That caused a problem when loading checkpoints with filenames like last-v3.ckpt using the code here

that looks for last.ckpt.