Correct typo in the staged path.

sdks/python/apache_beam/options/pipeline_options.py

@@ -1994,8 +1994,8 @@ def _add_argparse_args(cls, parser):
         help=(
             'Local path to a file. During job submission, the files will be '
             'staged in the staging area (--staging_location option) and then '
-            'workers will upload them to the worker specific staging location '
-            '(e.g. $SEMI_PERSISTENT_DIRECTORY/staged/ for portable runner.'))
+            'workers will download them to the worker-specific directory '
+            '$SEMI_PERSISTENT_DIRECTORY/staged/.'))
     parser.add_argument(
         '--prebuild_sdk_container_engine',
         help=(

website/www/site/content/en/documentation/sdks/python-pipeline-dependencies.md

@@ -109,10 +109,11 @@ Staged files are downloaded onto the worker:
 
       import my_module
 
-* **Data or configuration files**: If you staged non-Python files (such as a JSON config), they are downloaded to the staged files directory. You can locate this directory on the worker by reading the `SEMI_PERSISTENT_DIRECTORY` environment variable, or look for `/tmp/staged` which is the default location for staged files on containerized runners:
+* **Data or configuration files**: If you staged non-Python files (such as a JSON config), they are downloaded to the `$SEMI_PERSISTENT_DIRECTORY/staged` directory. The full path depends on the runner, so build it from the environment variable instead of hardcoding it.
 
       import os
-      staged_dir = os.environ.get('SEMI_PERSISTENT_DIRECTORY', '/tmp/staged')
+      staged_dir = os.path.join(
+          os.environ.get('SEMI_PERSISTENT_DIRECTORY', '/tmp'), 'staged')
       config_path = os.path.join(staged_dir, 'data_config.json')
 
 ### Importing Plugins on Worker Startup