[script] training on Multi-PIE: fixed the options (--light was missing)

bob/learn/pytorch/scripts/train_drgan_multipie.py

@@ -19,6 +19,7 @@ Options:
   -e, --epochs=<int>            The number of training epochs [default: 100] 
   -s, --sample=<int>            Save generated images at every 'sample' batch iteration [default: 100000000000] 
   -o, --output-dir=<path>       Dir to save the logs, models and images [default: ./drgan-light-mpie-casia/] 
+  -L, --light                   Use the DCGAN-based model instead of the original DR-GAN architecture 
   -g, --use-gpu                 Use the GPU 
   -S, --seed=<int>              The random seed [default: 3] 
   -v, --verbose                 Increase the verbosity (may appear multiple times).