-
Download preprocessed dataset
-
Extract data.zip in the root project directory
-
Run train.py from the root project directory
- Ensure that train.py
data_path
arg value points to the correct directory
- The preprocessed dataset contains some invalid data points. These occurences are handled in the dataloader by ignoring these datapoints. If a batch size of 64 is set, and some batches contain less than 64, this is why.