[Doc] Fix documentation in dgl.multiprocessing namespace (#3929)

* fix docs

* remove

* oh

* fix

docs/source/api/python/dgl.multiprocessing.rst

@@ -17,6 +17,5 @@ In addition, if your backend is PyTorch, this module will also be compatible wit
 .. autosummary::
     :toctree: ../../generated/
 
-    spawn
     call_once_and_share
     shared_tensor

python/dgl/multiprocessing/__init__.py

@@ -9,7 +9,7 @@ from .. import backend as F
 if F.get_preferred_backend() == 'pytorch':
     # Wrap around torch.multiprocessing...
     from torch.multiprocessing import *
-    # ... and override the Process initializer and spawn function.
+    # ... and override the Process initializer.
     from .pytorch import *
 else:
     # Just import multiprocessing module.

python/dgl/multiprocessing/pytorch.py

@@ -44,10 +44,8 @@ def _get_shared_mem_name(id_):
     return "shared" + str(id_)
 
 def call_once_and_share(func, shape, dtype, rank=0):
-    """Invoke the function in a single process of the process group spawned by
-    :func:`spawn`, and share the result to other processes.
-
-    Requires the subprocesses to be spawned with :func:`dgl.multiprocessing.pytorch.spawn`.
+    """Invoke the function in a single process of the PyTorch distributed process group,
+    and share the result with other processes.
 
     Parameters
     ----------
@@ -89,7 +87,7 @@ def call_once_and_share(func, shape, dtype, rank=0):
 
 def shared_tensor(shape, dtype=torch.float32):
     """Create a tensor in shared memory accessible by all processes within the same
-    ``torch.distsributed`` process group.
+    ``torch.distributed`` process group.
 
     The content is uninitialized.