misc: fix pr issues

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Signed-off-by: specture724 <149605198+specture724@users.noreply.github.com>

Author: specture724

checkpoint_engine/ps.py

@@ -1265,11 +1265,12 @@ def _update_per_bucket(
                 brank = bcast_rank_map[receiver_rank]
                 dist.broadcast(buffer_b, src=brank)
                 resp = socket.recv()
+                ret_code = torch.tensor(0, device="cuda")
                 if resp != b"":
                     logger.error(
                         f"[rank{self._rank}] receive error response from rank {receiver_rank} for bucket {gidx} in checkpoint {checkpoint_name}"
                     )
-                    ret_code = torch.tensor(1, device="cuda")
+                    ret_code.fill_(1)
                 dist.all_reduce(ret_code, op=dist.ReduceOp.SUM)
                 torch.cuda.synchronize()
                 if ret_code.item() != 0:

tests/test_update.py

@@ -46,6 +46,7 @@ def checker_proc_with_error(
 ):
     torch.cuda.set_device(rank)
     named_tensors = {name: tensor.cuda() for name, tensor in named_tensors.items()}
+    _ = named_tensors
     _zmq_ctx = zmq.Context()
 
     def trigger_error(socket_paths: list[tuple[str, str]]):
@@ -59,7 +60,7 @@ def trigger_error(socket_paths: list[tuple[str, str]]):
         )
 
     def error_run(weights: list[tuple[str, torch.Tensor]]):
-        weights = weights  # Do some fake processing
+        _ = weights  # Do some fake processing
         time.sleep(random.uniform(0.1, 0.5))
         if rank == 0:
             raise RuntimeError("Intentional Error for testing.")
@@ -207,7 +208,6 @@ def test_update(test_name: str, rank_list: list[list[int]] | None):
         sys.exit(1)
     assert len(sys.argv) > 2
     test_type = sys.argv[1]
-    world_size = get_world_size()
     rank_list = json.loads(sys.argv[2])
     if test_type == "test_no_error" or test_type == "long_test_no_error":
         run(checker_proc, rank_list, need_error=False)