the file path issue of different deploy modes in spark

Question

I am writing a simlpe spark application using python. It seems that if you save a dataframe in the path "tmp/df", you can access the files using "hadoop fs -g