Pytorch=1.1.0
Ubuntu (moviepy may not work properly on Windows)
Download the original dataset from Berkeley Server and extract
Run data_converter.py to convert the TFRecord to .pkl python data_converter.py --input_path <Where your bair dataset is>
Run python train.py --model_name cdna --horizon 10 --epoch 10
to train the model.
model_name
can be choose from etd
etds
etdm
cdna
.
You can use tensorboard to visualize the results tensorboard --logdir=runs
Run python test.py --model_name cdna --horizon 20 --load_point 10
The result will be generated in model/bair/<model_name>_10/test_<horizon>/