Revert "torchx/runner: log events to torch.monitor (#379)" (#536)

Summary:
This reverts commit 4b989d5.

Importing torch.monitor adds a pretty large dependency on torch which hugely slows down load time in OSS.

This accounts for about 340ms of load time.

Pull Request resolved: #536

Test Plan: This isn't being used anywhere AFAIK so should be safe to land. Internal has a custom handler.

Reviewed By: priyaramani, kurman

Differential Revision: D37432126

Pulled By: d4l3k

fbshipit-source-id: f38a89a7a4dbbd69e9ba43c130bd31b084c0bb00

Author: d4l3k

.pyre_configuration

@@ -8,7 +8,6 @@
   ],
   "site_package_search_strategy": "all",
   "source_directories": [
-    "typestubs",
     "."
   ],
   "strict": true,

torchx/cli/cmd_run.py

@@ -116,16 +116,16 @@ def __init__(self) -> None:
         self._subparser: Optional[argparse.ArgumentParser] = None
 
     def add_arguments(self, subparser: argparse.ArgumentParser) -> None:
-        scheduler_names = get_scheduler_factories().keys()
+        # scheduler_names = get_scheduler_factories().keys()
         self._subparser = subparser
         subparser.add_argument(
             "-s",
             "--scheduler",
             type=str,
-            default=get_default_scheduler_name(),
-            choices=list(scheduler_names),
+            # default=get_default_scheduler_name(),
+            # choices=list(scheduler_names),
             action=torchxconfig_run,
-            help=f"Name of the scheduler to use. One of: [{','.join(scheduler_names)}]",
+            # help=f"Name of the scheduler to use. One of: [{','.join(scheduler_names)}]",
         )
         subparser.add_argument(
             "-cfg",

torchx/runner/events/__init__.py

@@ -59,16 +59,6 @@ def _get_or_create_logger(destination: str = "null") -> logging.Logger:
 def record(event: TorchxEvent, destination: str = "null") -> None:
     _get_or_create_logger(destination).info(event.serialize())
 
-    if destination != "console":
-        # if using torch>1.11 log the event to torch.monitor
-        try:
-            from torch import monitor
-
-            # pyre-fixme[16]: Module `monitor` has no attribute `log_event`.
-            monitor.log_event(event.to_monitor_event())
-        except ImportError:
-            pass
-
 
 class log_event:
     """

torchx/runner/events/api.py

@@ -7,12 +7,8 @@
 
 import json
 from dataclasses import asdict, dataclass
-from datetime import datetime
 from enum import Enum
-from typing import Optional, TYPE_CHECKING, Union
-
-if TYPE_CHECKING:
-    from torch import monitor
+from typing import Optional, Union
 
 
 class SourceType(str, Enum):
@@ -64,14 +60,3 @@ def deserialize(data: Union[str, "TorchxEvent"]) -> "TorchxEvent":
 
     def serialize(self) -> str:
         return json.dumps(asdict(self))
-
-    # pyre-fixme[11]: Annotation `Event` is not defined as a type.
-    def to_monitor_event(self) -> "monitor.Event":
-        from torch import monitor
-
-        # pyre-fixme[16]: Module `monitor` has no attribute `Event`.
-        return monitor.Event(
-            name="torch.runner.Event",
-            timestamp=datetime.now(),
-            data={k: v for k, v in self.__dict__.items() if v is not None},
-        )

torchx/runner/events/test/lib_test.py

@@ -8,24 +8,15 @@
 import json
 import logging
 import unittest
-from typing import List
 from unittest.mock import MagicMock, patch
 
 from torchx.runner.events import (
     _get_or_create_logger,
     log_event,
-    record,
     SourceType,
     TorchxEvent,
 )
 
-try:
-    from torch import monitor
-
-    SKIP_MONITOR: bool = False
-except ImportError:
-    SKIP_MONITOR: bool = True
-
 
 class TorchxEventLibTest(unittest.TestCase):
     def assert_event(
@@ -66,55 +57,6 @@ def test_event_deser(self) -> None:
         deser_event = TorchxEvent.deserialize(json_event)
         self.assert_event(event, deser_event)
 
-    @unittest.skipIf(SKIP_MONITOR, "no torch.monitor available")
-    def test_monitor(self) -> None:
-        event = TorchxEvent(
-            session="test_session",
-            scheduler="test_scheduler",
-            api="test_api",
-            source=SourceType.EXTERNAL,
-        )
-        monitor_event = event.to_monitor_event()
-        self.assertEqual(
-            monitor_event.data,
-            {
-                "session": "test_session",
-                "scheduler": "test_scheduler",
-                "api": "test_api",
-                "source": "EXTERNAL",
-            },
-        )
-        self.assertEqual(monitor_event.name, "torch.runner.Event")
-
-    @unittest.skipIf(SKIP_MONITOR, "no torch.monitor available")
-    @patch("torchx.runner.events._get_or_create_logger")
-    def test_monitor_record(self, get_logging_handler: MagicMock) -> None:
-        event = TorchxEvent(
-            session="test_session",
-            scheduler="test_scheduler",
-            api="test_api",
-            source=SourceType.EXTERNAL,
-        )
-        # pyre-fixme[11]: Annotation `Event` is not defined as a type.
-        events: List[monitor.Event] = []
-
-        def handler(e: monitor.Event) -> None:
-            events.append(e)
-
-        # pyre-fixme[16]: Module `monitor` has no attribute `register_event_handler`.
-        handle = monitor.register_event_handler(handler)
-
-        try:
-            record(event)
-        finally:
-            # pyre-fixme[16]: Module `monitor` has no attribute
-            #  `unregister_event_handler`.
-            monitor.unregister_event_handler(handle)
-
-        self.assertEqual(get_logging_handler.call_count, 1)
-        self.assertEqual(len(events), 1)
-        self.assertEqual(events[0].data["session"], "test_session")
-
 
 @patch("torchx.runner.events.record")
 class LogEventTest(unittest.TestCase):

torchx/schedulers/local_scheduler.py

@@ -1063,7 +1063,7 @@ def __next__(self) -> str:
                     return line
 
 
-def create_cwd_scheduler(session_name: str, **kwargs: Any) -> LocalScheduler:
+def create_scheduler(session_name: str, **kwargs: Any) -> LocalScheduler:
     return LocalScheduler(
         session_name=session_name,
         cache_size=kwargs.get("cache_size", 100),