Merge pull request #120 from rapidsai/branch-24.02

Forward-merge branch-24.02 to branch-24.04

python/pylibwholegraph/pylibwholegraph/torch/comm.py

@@ -32,6 +32,19 @@
 all_comm_local_size = 1
 
 
+def reset_communicators():
+    global all_comm_world_rank, all_comm_world_size, all_comm_local_rank, all_comm_local_size
+    global global_communicators, local_node_communicator, local_device_communicator
+    global_communicators = {}
+    local_node_communicator = None
+    local_device_communicator = None
+
+    all_comm_world_rank = 0
+    all_comm_world_size = 1
+    all_comm_local_rank = 0
+    all_comm_local_size = 1
+
+
 def set_world_info(world_rank: int, world_size: int, local_rank: int, local_size: int):
     """
     Set the global world's information. This is used for create common used communicators, like local node communicator,

python/pylibwholegraph/pylibwholegraph/torch/initialize.py

@@ -15,7 +15,7 @@
 import torch
 import torch.utils.dlpack
 import pylibwholegraph.binding.wholememory_binding as wmb
-from .comm import set_world_info, get_global_communicator, get_local_node_communicator
+from .comm import set_world_info, get_global_communicator, get_local_node_communicator, reset_communicators
 
 
 def init(world_rank: int, world_size: int, local_rank: int, local_size: int):
@@ -73,3 +73,5 @@ def finalize():
     :return: None
     """
     wmb.finalize()
+    reset_communicators()
+    torch.distributed.destroy_process_group() if torch.distributed.is_initialized() else None