diff --git a/docs/source/examples/fully_featured_trainer.py b/docs/source/examples/fully_featured_trainer.py
index 999b1c45cf..c6cb3833fb 100644
--- a/docs/source/examples/fully_featured_trainer.py
+++ b/docs/source/examples/fully_featured_trainer.py
@@ -169,29 +169,30 @@ if __name__ == '__main__':
 
     # format GPU layout
     os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID"
-    gpu_ids = hyperparams.gpus.split(';')
 
     # RUN TRAINING
     if hyperparams.on_cluster:
-        # Gets called when running via HPC cluster
+        # run on HPC cluster
         print('RUNNING ON SLURM CLUSTER')
+        gpu_ids = hyperparams.gpus.split(';')
         os.environ["CUDA_VISIBLE_DEVICES"] = ','.join(gpu_ids)
         optimize_on_cluster(hyperparams)
 
+    elif hyperparams.gpus is None:
+        # run on cpu
+        print('RUNNING ON CPU')
+        main(hyperparams, None, None)
+
     elif hyperparams.single_run_gpu:
         # run on 1 gpu
+        gpu_ids = hyperparams.gpus.split(';')
         print(f'RUNNING 1 TRIAL ON GPU. gpu: {gpu_ids[0]}')
         os.environ["CUDA_VISIBLE_DEVICES"] = gpu_ids[0]
         main(hyperparams, None, None)
 
-    elif hyperparams.local or hyperparams.single_run:
-        # run 1 trial but on CPU
-        os.environ["CUDA_VISIBLE_DEVICES"] = '0'
-        print('RUNNING LOCALLY')
-        main(hyperparams, None, None)
-
     else:
         # multiple GPUs on same machine
+        gpu_ids = hyperparams.gpus.split(';')
         print(f'RUNNING MULTI GPU. GPU ids: {gpu_ids}')
         hyperparams.optimize_parallel_gpu(
             main_local,