Basic example of a TensorFlow project. The code runs over TensorFlow 1.10.
We can first we test the base model without test dataset in data.txt:
python main.py --data data.txt test
We train the model with the synthetic generated datset:
python main.py train
Finally we test our trained model with the test dataset in data.txt:
python main.py --data data.txt test
We can se the results in TensorBoard:
tensorboard --logdir=./outputs
Training can be launched with different set of parameters, here a complete example:
python main.py --log_steps 1 --outputs_dir ./output --num_neurons 4,4,2 --epochs 10 --batch_size 2 --learning_rate 0.1 --learning_rate_decay 0.1 --train_shuffle_size 100 --max_time "2hours 30min" --random_seed 42 train
Type python main.py -h for help about the parameters