search

microsoft / Recursive-Cascaded-Networks

[ICCV 2019] Recursive Cascaded Networks for Unsupervised Medical Image Registration
https://arxiv.org/abs/1907.12353
MIT License
362 stars 87 forks source link

How to continue the training on the saved model? #23

Closed CGC1031 closed 4 years ago

CGC1031 commented 4 years ago

According to my computer's hardware configuration, a full training session with 20,000 rounds takes a long time.How do you build on the model that you trained before?For example, I have trained 10,000 rounds and saved them. How can I continue the training on the saved model? Do you have any good Suggestions? Looking forward to your reply！

zsyzzsoft commented 4 years ago

Use the '-c' argument!

CGC1031 commented 4 years ago

Thank you very much,this problem has been solved successfully.

CGC1031 commented 4 years ago

May I ask what is the meaning of 'Total time' and 'data' ? Looking forward to your reply.Thank you very much. QQ图片20200518222729

zsyzzsoft commented 4 years ago

"Total time" is the measured time of a step; "data" is the percentage of data loading time over the "Total time", expected to be small.

CGC1031 commented 4 years ago

Thank you for your reply , but I'm still a little confused. As shown in the figure below , 200 steps take 2 minutes and 44 seconds , for a total of 164 seconds. While "Total time" is the measure time of one step equals 0.29 second , so 200*0.29 = 58 seconds . Why are the calculated times unequal ? 微信图片_20200520102110

CGC1031 commented 4 years ago

This is picture.Thank you. QQ图片20200520103249

zsyzzsoft commented 4 years ago

This is only for monitoring purpose, and the "Total time" only measures the time of some step. You may modify the code to get more accurate values.

CGC1031 commented 4 years ago

OK，thank you very much.