Thank you for your great work. I have a question about evaluation. I noticed that checkpoints are regularly saved during the training process, and there are multiple checkpoints after the training process is over. When evaluating, you only use the last checkpoint to report performance, or use all checkpoints and report the best performance?
Thank you for your great work. I have a question about evaluation. I noticed that checkpoints are regularly saved during the training process, and there are multiple checkpoints after the training process is over. When evaluating, you only use the last checkpoint to report performance, or use all checkpoints and report the best performance?