/arnn

Primary LanguagePython

arnn

Train the model

To train the model, run

$ python main.py --train --model_name <name of your model>

This trains a model with default parameters:

  • Training epochs: --epoch 150
  • Mini batch size: --batch_size 64
  • GRU units: --rnn 20
  • Length of sequence for training: --len 40
  • Dropout rate: --drop 0.001

Test the model

Test on a specific sequence

To test on a specific sequence in the test data and outputting a prediction sequence of your choice, you can use the --test_sample and --test_steps flags.

For instance, running the command

$ python main.py --test --model_name <name of your model> --test_sample 0:30 --test_steps 10

will feed the model preprocessed samples from row index 0 to 30 as an input sequence and the it will predict the values that are 10 steps ahead.

Visualize the predicted values

--test_sample is usually used with --plot_sample to visualize the predicted samples against the target values.

$ python main.py --test --model_name <name of your model> --test_sample 0:30 --test_steps 10 --plot_sample