expose memory_fraction and gpu_process docs (#51372)

Summary:
Per title

Pull Request resolved: #51372

Reviewed By: mruberry

Differential Revision: D26157787

Pulled By: ngimel

fbshipit-source-id: 97eac5f12881a2bf62c251f6f7eaf65fdbe34056

docs/source/cuda.rst

@@ -44,6 +44,7 @@ Streams and events
 Memory management
 -----------------
 .. autofunction:: empty_cache
+.. autofunction:: list_gpu_processes
 .. autofunction:: memory_stats
 .. autofunction:: memory_summary
 .. autofunction:: memory_snapshot
@@ -52,6 +53,7 @@ Memory management
 .. autofunction:: reset_max_memory_allocated
 .. autofunction:: memory_reserved
 .. autofunction:: max_memory_reserved
+.. autofunction:: set_per_process_memory_fraction
 .. FIXME The following doesn't seem to exist. Is it supposed to?
    https://github.com/pytorch/pytorch/issues/27785
    .. autofunction:: reset_max_memory_reserved