Currently, rerunning an inference pipeline will save data in the exact same spot as a previous run. This will overrun previous models, requiring the user to move the old version manually without rerunning. It would be nice if we had an easy-to-load configuration parameter which provided a name, prefix, or custom directory to our save files to prevent them from overwriting. By default, this could be automatically configured from the type of model training and the time of run.
Currently, rerunning an inference pipeline will save data in the exact same spot as a previous run. This will overrun previous models, requiring the user to move the old version manually without rerunning. It would be nice if we had an easy-to-load configuration parameter which provided a name, prefix, or custom directory to our save files to prevent them from overwriting. By default, this could be automatically configured from the type of model training and the time of run.