Keeping the state after loading from a checkpoint

2 ビュー (過去 30 日間)
John Smith
John Smith 2023 年 3 月 2 日
コメント済み: John Smith 2023 年 3 月 6 日
I'm running an experiment using Deep Learning Toolbox. I'd like to resume execution after restarting from a checkpoint. However, the results aren't the same after every restart eventhough I save the global random generator state using s = rng(); before checkpointing and restoring it back with rng(s); after the loading from the checkpoint.
If I restart Matlab and then restart from the checkpoint then the results are the same.
What else do I need to save s.t. I can restart from a checkpoint and obtain the same results?
  1 件のコメント
John Smith
John Smith 2023 年 3 月 6 日
After more digging, the problem is that Matlab resets the datastore/minibatchqueue after reloading from a checkpoint rather than continuing from the checkpoint.
Is this a bug or an (undocumented) feature?


回答 (0 件)


Find more on Deep Learning Toolbox in Help Center and File Exchange

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by