use 'python' rather than 'tensorflow' prefix for S3 types; additional checks for python availability before executing various methods