TensorFlow template

We have 3 .py files and 2 folders which are described below.

• main.py

  • Define the whole process.

• model.py

  • Define the architecture of the model.
  • methods:
    • build: build the model.
    • sub_model: define the architecture of the model.
    • predict: predict the labels according to the model.
    • loss: define the loss of the model.

• trainer.py

  • Define how to train the model.
  • methods:
    • train: train the model.
    • result: show the result.
  • arguments:
    • sess: tf.sess declared outside Trainer
    • epoch: the number of epoch we need to train
    • print_every: how often we print the result

• logs

  • Record the training process
  • Usage: python -m tensorboard.main --logdir=path/to/logs

• result

  • The visualized result.