[torch] Set nproc_per_node to 1 (#61552) (#64791)

Summary: Pull Request resolved: #61552 Set `nproc_per_node` to 1 Test Plan: unittests Reviewed By: cbalioglu Differential Revision: D29667056 fbshipit-source-id: 6601f66fec5e018c7737d909f8c71642451abb29 Co-authored-by: Aliaksandr Ivanou <[email protected]>
pytorch · Sep 10, 2021 · 9cc96ad · 9cc96ad
1 parent 7a9f23c
commit 9cc96ad
Show file tree

Hide file tree

Showing 2 changed files with 31 additions and 1 deletion.
diff --git a/test/distributed/launcher/run_test.py b/test/distributed/launcher/run_test.py
@@ -169,6 +169,35 @@ def test_launch_user_script_bash(self):
             {str(i) for i in range(world_size)}, set(os.listdir(self.test_dir))
         )
 
+    @unittest.skipIf(TEST_WITH_ASAN or TEST_WITH_TSAN, "test incompatible with tsan")
+    def test_launch_user_script_default_nproc(self):
+        run_id = str(uuid.uuid4().int)
+        nnodes = 1
+        world_size = 1
+        args = [
+            f"--nnodes={nnodes}",
+            "--rdzv_backend=etcd",
+            f"--rdzv_endpoint={self._etcd_endpoint}",
+            f"--rdzv_id={run_id}",
+            "--monitor_interval=1",
+            "--start_method=fork",
+            "--no_python",
+        ]
+
+        script_args = [path("bin/test_script.sh"), f"{self.test_dir}"]
+
+        with self.assertRaises(ValueError):
+            # --no_python cannot be used with --module
+            launch.main(args + ["--module"] + script_args)
+
+        launch.main(args + script_args)
+
+        # make sure all the workers ran
+        # each worker touches a file with its global rank as the name
+        self.assertSetEqual(
+            {str(i) for i in range(world_size)}, set(os.listdir(self.test_dir))
+        )
+
     @unittest.skipIf(TEST_WITH_ASAN or TEST_WITH_TSAN, "test incompatible with tsan")
     def test_launch_with_env_vars(self):
         run_id = str(uuid.uuid4().int)

diff --git a/torch/distributed/run.py b/torch/distributed/run.py
@@ -240,6 +240,7 @@ def train():
 import torch
 from torch.distributed.argparse_util import check_env, env
 from torch.distributed.elastic.multiprocessing import Std
+from torch.distributed.elastic.multiprocessing.errors import record
 from torch.distributed.elastic.rendezvous.utils import _parse_rendezvous_config
 from torch.distributed.elastic.utils import macros
 from torch.distributed.elastic.utils.logging import get_logger
@@ -269,7 +270,7 @@ def get_args_parser() -> ArgumentParser:
         "--nproc_per_node",
         action=env,
         type=str,
-        default="auto",
+        default="1",
         help="Number of workers per node; supported values: [auto, cpu, gpu, int].",
     )