Closed DimanChauncey closed 8 months ago
According to the official example, each task needs to be resampled randomly at the beginning. Is there a checkpoint mechanism? Or can I read the results of previous training and give parameters based on that for the next iteration? What should I do?
Hi @DimanChauncey, you may use advisor.save_json(filename) and advisor.load_json(filename) to save and load existing observations.
advisor.save_json(filename)
advisor.load_json(filename)
According to the official example, each task needs to be resampled randomly at the beginning. Is there a checkpoint mechanism? Or can I read the results of previous training and give parameters based on that for the next iteration? What should I do?