torchmonarch-nightly 2025.7.30__cp311-cp311-manylinux2014_x86_64.whl → 2025.7.31__cp311-cp311-manylinux2014_x86_64.whl

tests/test_allocator.py

@@ -104,7 +104,9 @@ class TestActor(Actor):
 
 @contextlib.contextmanager
 def remote_process_allocator(
-    addr: Optional[str] = None, timeout: Optional[int] = None
+    addr: Optional[str] = None,
+    timeout: Optional[int] = None,
+    envs: Optional[dict[str, str]] = None,
 ) -> Generator[str, None, None]:
     """Start a remote process allocator on addr. If timeout is not None, have it
     timeout after that many seconds if no messages come in"""
@@ -120,16 +122,19 @@ def remote_process_allocator(
         if timeout is not None:
             args.append(f"--timeout-sec={timeout}")
 
+        env = {
+            # prefix PATH with this test module's directory to
+            # give 'process_allocator' and 'monarch_bootstrap' binary resources
+            # in this test module's directory precedence over the installed ones
+            # useful in BUCK where these binaries are added as 'resources' of this test target
+            "PATH": f"{package_path}:{os.getenv('PATH', '')}",
+            "RUST_LOG": "debug",
+        }
+        if envs:
+            env.update(envs)
         process_allocator = subprocess.Popen(
             args=args,
-            env={
-                # prefix PATH with this test module's directory to
-                # give 'process_allocator' and 'monarch_bootstrap' binary resources
-                # in this test module's directory precedence over the installed ones
-                # useful in BUCK where these binaries are added as 'resources' of this test target
-                "PATH": f"{package_path}:{os.getenv('PATH', '')}",
-                "RUST_LOG": "debug",
-            },
+            env=env,
         )
         try:
             yield addr
@@ -233,6 +238,26 @@ class TestRemoteAllocator(unittest.IsolatedAsyncioTestCase):
         computed_world_sizes = {p.rank: v for p, v in list(computed.flatten("rank"))}
         self.assertDictEqual(expected_world_sizes, computed_world_sizes)
 
+    async def test_allocate_failure_message(self) -> None:
+        spec = AllocSpec(AllocConstraints(), host=2, gpu=4)
+
+        with self.assertRaisesRegex(
+            Exception,
+            r"exited with code 1: Traceback \(most recent call last\).*",
+        ):
+            with remote_process_allocator(
+                envs={"MONARCH_ERROR_DURING_BOOTSTRAP_FOR_TESTING": "1"}
+            ) as host1, remote_process_allocator(
+                envs={"MONARCH_ERROR_DURING_BOOTSTRAP_FOR_TESTING": "1"}
+            ) as host2:
+                allocator = RemoteAllocator(
+                    world_id="test_remote_allocator",
+                    initializer=StaticRemoteAllocInitializer(host1, host2),
+                    heartbeat_interval=_100_MILLISECONDS,
+                )
+                alloc = await allocator.allocate(spec)
+                await ProcMesh.from_alloc(alloc)
+
     async def test_call_allocate_twice(self) -> None:
         class DeletingAllocInitializer(StaticRemoteAllocInitializer):
             """test initializer that removes the last address from the list each time initialize_alloc() is called

{torchmonarch_nightly-2025.7.30.dist-info → torchmonarch_nightly-2025.7.31.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: torchmonarch-nightly
-Version: 2025.7.30
+Version: 2025.7.31
 Summary: Monarch: Single controller library
 Author: Meta
 Author-email: oncall+monarch@xmail.facebook.com

{torchmonarch_nightly-2025.7.30.dist-info → torchmonarch_nightly-2025.7.31.dist-info}/RECORD

@@ -1,5 +1,5 @@
 monarch/__init__.py,sha256=mgKiyD1kxky-1pvhMlNfF4VmxWnhi-FSYZNFzkW1BEM,7052
-monarch/_rust_bindings.so,sha256=9nuRahXA9kYCHXLJBKIuGFGemXDe_lbvVflzMjmJVPc,50318152
+monarch/_rust_bindings.so,sha256=3bAHSTsQ7tHjGqEctI8uiIAvZDNSWYv-a8Q5ABi-9JI,50324720
 monarch/_testing.py,sha256=_3MYNMq-_0T1qXCj2vxrW13GlWGdUuVFMskQF2Gsw_o,7877
 monarch/actor_mesh.py,sha256=VtPU9syi_vUdwDSJJ639Z4Y_EcWZUScyoj0lQ88RQPs,421
 monarch/bootstrap_main.py,sha256=39OZpNMrfvvNJf-iwuNzgslzYA_ItaRPHfXGn_V74N0,524
@@ -8,7 +8,7 @@ monarch/fetch.py,sha256=JMxC8HdWMvpik0T4E6e-gfxvmNnOkA0ul4eo4R3Jg_o,1712
 monarch/gradient_generator.py,sha256=Rl3dmXGceTdCc1mYBg2JciR88ywGPnW7TVkL86KwqEA,6366
 monarch/memory.py,sha256=ol86dBhFAJqg78iF25-BuK0wuwj1onR8FIioZ_B0gjw,1377
 monarch/mesh_controller.py,sha256=R9ZnVV89wYva0QTAwOgHi_PkjYPEj_7_yF9810NHPak,14675
-monarch/monarch_controller,sha256=wN7745JWkXuivdYrtoCEnsYDXoEfgjMAJuRdguZ0lzs,24035224
+monarch/monarch_controller,sha256=bhnWJ9gmhqYjtG-SDKN6K7yl93k6R3HAIwnA6HTbkB0,24028112
 monarch/notebook.py,sha256=zu9MKDFKf1-rCM2TqFSRJjMBeiWuKcJSyUFLvoZRQzs,25949
 monarch/opaque_module.py,sha256=jCcg0DjbcEVXA9WNG0NhUzGteLHOJLTZEBvrIYJIAns,10436
 monarch/opaque_object.py,sha256=x1LoX6RIMGh4ux52xIfhPgoh6PhZHdkf9bMccHW3DW0,2808
@@ -89,7 +89,7 @@ monarch/controller/rust_backend/__init__.py,sha256=J8qjUOysmcMAek2KFN13mViOXZxTY
 monarch/controller/rust_backend/controller.py,sha256=8IYnVUiqEVKO9rGL3vKqcCSAhWJG1bYYQ0MoaMqsp78,9521
 monarch/gradient/__init__.py,sha256=kqmzwt16mMpk0M3GhpgP_f7da4DGnaV9chDzbt66k4Q,308
 monarch/gradient/_gradient_generator.pyi,sha256=6cX0UxaDt9NAlwgIhTgnweqGOf6qRhHiGnUzSWNCxdU,630
-monarch/gradient/_gradient_generator.so,sha256=VyBW9SsE1IJ5iiiq1Su0BgTR7vDBnbaRQj7yjAtBPUs,11531728
+monarch/gradient/_gradient_generator.so,sha256=fVFAMTvuSABPN6oBZ-J-Tcf0spk1WgTsbaSmCz4pALU,11531728
 monarch/parallel/__init__.py,sha256=6920kIkhiX7AiyjYvyc1ad8ccP-bStJJ1sS5KkeN2P0,352
 monarch/parallel/pipelining/__init__.py,sha256=J8qjUOysmcMAek2KFN13mViOXZxTYc5vCrF02t3VuFU,223
 monarch/parallel/pipelining/runtime.py,sha256=KK8TG1gUYEzSsquiZoPTWGSIC74mlncD7cYknKxfb3c,32470
@@ -147,7 +147,7 @@ tests/sleep_binary.py,sha256=XfLYaAfwm9xgzM-svs8fhAeFhwYIg6SyVEnx4e6wbUw,1009
 tests/test_actor_error.py,sha256=U7QL1jRn-YpS-o62imt7HFLPtaSbwMBu9xpD09Mb-Bc,20875
 tests/test_actor_shape.py,sha256=ph-RC9sMNHWptZOCwQqMfG4lIUEzhp_pEnfhITeYdHM,4533
 tests/test_alloc.py,sha256=IW7yJSaKxhOYc8SJtFyREakDUwiKWq9M0CGgYyBYHoc,743
-tests/test_allocator.py,sha256=4LcUB4QRNGDp0qBWAyLM6ektmoxpO922f-NcHZziJ_w,28762
+tests/test_allocator.py,sha256=DpQhygQ4jB19g-aY-BFR61J-gYbG-hDlTNPVuQ4Fmn0,29730
 tests/test_coalescing.py,sha256=JZ4YgQNlWWs7N-Z8KCCXQPANcuyyXEKjeHIXYbPnQhk,15606
 tests/test_controller.py,sha256=CIMb-ApmBcBj1eCqccDUAbVyyJWMGooAha5gQk0AoeY,31452
 tests/test_debugger.py,sha256=9opgQXCBuZ1Z-7uOKI-FuGB0jLbLLilmWQKq0sE-dgQ,21950
@@ -170,9 +170,9 @@ tests/simulator/test_profiling.py,sha256=TGYCfzTLdkpIwnOuO6KApprmrgPIRQe60KRX3wk
 tests/simulator/test_simulator.py,sha256=LO8lA0ssY-OGEBL5ipEu74f97Y765TEwfUOv-DtIptM,14568
 tests/simulator/test_task.py,sha256=ipqBDuDAysuo1xOB9S5psaFvwe6VATD43IovCTSs0t4,2327
 tests/simulator/test_worker.py,sha256=QrWWIJ3HDgDLkBPRc2mwYPlOQoXQcj1qRfc0WUfKkFY,3507
-torchmonarch_nightly-2025.7.30.dist-info/licenses/LICENSE,sha256=e0Eotbf_rHOYPuEUlppIbvwy4SN98CZnl_hqwvbDA4Q,1530
-torchmonarch_nightly-2025.7.30.dist-info/METADATA,sha256=TLI6C3RGzWoCq2Ual0X4CEz_oUaMFvHdAxIsAI6yxaI,3852
-torchmonarch_nightly-2025.7.30.dist-info/WHEEL,sha256=JC9FVdjbTDi9l3EyrqUd11CgmN9LkBi1g5dFHayafwA,104
-torchmonarch_nightly-2025.7.30.dist-info/entry_points.txt,sha256=60QVSpYVzkzS4iDOiLp0fsLxVp47X3J2l3v7W-59LMo,117
-torchmonarch_nightly-2025.7.30.dist-info/top_level.txt,sha256=E-ZssZzyM17glpVrh-S9--qJ-w9p2EjuYOuNw9tQ4Eg,33
-torchmonarch_nightly-2025.7.30.dist-info/RECORD,,
+torchmonarch_nightly-2025.7.31.dist-info/licenses/LICENSE,sha256=e0Eotbf_rHOYPuEUlppIbvwy4SN98CZnl_hqwvbDA4Q,1530
+torchmonarch_nightly-2025.7.31.dist-info/METADATA,sha256=V4U9RkS-KxbvPEmkrhGYp3WRtTXluOVzAl9ktisydYw,3852
+torchmonarch_nightly-2025.7.31.dist-info/WHEEL,sha256=JC9FVdjbTDi9l3EyrqUd11CgmN9LkBi1g5dFHayafwA,104
+torchmonarch_nightly-2025.7.31.dist-info/entry_points.txt,sha256=60QVSpYVzkzS4iDOiLp0fsLxVp47X3J2l3v7W-59LMo,117
+torchmonarch_nightly-2025.7.31.dist-info/top_level.txt,sha256=E-ZssZzyM17glpVrh-S9--qJ-w9p2EjuYOuNw9tQ4Eg,33
+torchmonarch_nightly-2025.7.31.dist-info/RECORD,,