search

dhlab-epfl / dhSegment

Generic framework for historical document processing
https://dhlab-epfl.github.com/dhSegment
GNU General Public License v3.0
373 stars 116 forks source link

Model is not exported after training #20

Closed stefan-it closed 5 years ago

stefan-it commented 6 years ago

Hi :)

I trained a model from scratch for ~ 10 epochs. Unfortunately, the model is not exported (the export folder is empty).

Is there any workaround to create an exported model from the checkpoints?

Thanks in advance :)

solivr commented 5 years ago

Hi,

There should be a model exported at the end of training, no matter how many epochs... Have you tried training for more than 10 epochs ? You can continue training you model using python train.py with config.json restore_model=True. You can also try to change in config.json the parameter evaluate_every_epoch to a lower value, so that your model gets exported after the evaluation.

Otherwise you can use tf.Estimator to export your model (see the doc here).

solivr commented 5 years ago

Closing because of inactivity.