Checkpoint for trained models

[parthnatekar]

Hi, where can I find the checkpoints for the trained models.

According to documentation here, you need to load the graph using the saved GraphDef and then restore the checkpoints. I can't seem to find where the checkpoints are being restored.

I need to run the evaluate.py script on the trained models.

[jonaballe]

Hi there, checkpoints are not necessary. The metagraphs that we publish (see tfci.py for code to load/run them) have all the variables baked in as constants.