search

Zhangjinso / PISE

123 stars 28 forks source link

How to continue training after interruption? #9

Closed Ha0Tang closed 3 years ago

Zhangjinso commented 3 years ago

You can refer to the options used in this work. With --continue_train and specific --which_iter (default is latest), you can continue training your model.

Ha0Tang commented 3 years ago

@Zhangjinso --continue_train will start from epoch 0?

Zhangjinso commented 3 years ago

It is strange. If you set --continue_train, the default option of which_iter is latest (means the latest model in your specific folder). You can refer to line 43--45 to check this process.