Automatic Logger Level Update (pytorch#1206)

* automatic logger level update

* add config log level unit test

* fix lint

* fix import errors

* revert init_logging move

* fix import error

* refactor out logging

* move logging back to util

pytest.ini

@@ -1,5 +1,5 @@
 [pytest]
 testpaths =
     test
-log_cli = True
+log_cli = False
 log_cli_level = INFO

test/test_misc.py

@@ -4,6 +4,7 @@
 import dataclasses
 import dis
 import enum
+import logging
 import math
 import os
 import sys
@@ -2330,6 +2331,21 @@ def f3(x):
             f3(torch.ones(6))
         self.assertEqual(cnt.frame_count, 0)
 
+    def test_config_log_level(self):
+        @torchdynamo.optimize("eager")
+        def fn(a, b):
+            return a + b
+
+        with self.assertLogs(logger="torchdynamo", level=logging.DEBUG) as log:
+            torchdynamo.config.log_level = logging.DEBUG
+            fn(torch.randn(10), torch.randn(10))
+            cur_len = len(log)
+            self.assertGreater(cur_len, 0)
+
+            torchdynamo.config.log_level = logging.WARNING
+            fn(torch.randn(10), torch.randn(10))
+            self.assertEqual(cur_len, len(log))
+
 
 class TestTracer(JitTestCase):
     def test_jit_save(self):

test/test_recompile_ux.py

@@ -85,7 +85,7 @@ def model(input):
         with unittest.mock.patch.object(
             torchdynamo.config, "cache_size_limit", expected_recompiles
         ):
-            with self.assertLogs(level="WARNING") as logs:
+            with self.assertLogs(logger="torchdynamo", level="WARNING") as logs:
                 for _ in range(10):
                     bsz = torch.randint(low=0, high=1000, size=())
                     x = torch.randn((bsz, 3, 4))
@@ -152,7 +152,7 @@ def cache_fail_test(cached_input, missed_input, expected_failure):
             # warmup
             opt_func(cached_input)
 
-            with self.assertLogs(level="WARNING") as logs:
+            with self.assertLogs(logger="torchdynamo", level="WARNING") as logs:
                 opt_func = torchdynamo.optimize("eager")(func)
                 opt_func(missed_input)
             self.assert_single_log_contains(logs, expected_failure)
@@ -190,7 +190,7 @@ def func(a, b):
         # warmup
         opt_func(a, b)
 
-        with self.assertLogs(level="WARNING") as logs:
+        with self.assertLogs(logger="torchdynamo", level="WARNING") as logs:
             opt_func = torchdynamo.optimize("eager")(func)
             opt_func(a, 1)
         self.assert_single_log_contains(

torchdynamo/config.py

@@ -7,6 +7,8 @@
 
 import torch
 
+import torchdynamo.utils
+
 try:
     import torch._prims
     import torch._refs
@@ -21,13 +23,15 @@
 # INFO print compiled functions + graphs
 # WARN print warnings (including graph breaks)
 # ERROR print exceptions (and what user code was being processed when it occurred)
+# NOTE: changing log_level will automatically update the levels of all torchdynamo loggers
 log_level = logging.WARNING
-# Verbose will print full stack traces on warnings and errors
-verbose = False
 
 # the name of a file to write the logs to
 log_file_name = None
 
+# Verbose will print full stack traces on warnings and errors
+verbose = False
+
 # verify the correctness of optimized backend
 verify_correctness = False
 
@@ -130,6 +134,9 @@ class _AccessLimitingConfig(ModuleType):
     def __setattr__(self, name, value):
         if name not in _allowed_config_names:
             raise AttributeError(f"{__name__}.{name} does not exist")
+        # automatically set logger level whenever config.log_level is modified
+        if name == "log_level":
+            torchdynamo.utils.set_loggers_level(value)
         return object.__setattr__(self, name, value)
 
 

torchdynamo/testing.py

@@ -1,7 +1,6 @@
 import contextlib
 import dis
 import functools
-import logging
 import os.path
 import types
 import unittest
@@ -194,7 +193,6 @@ def tearDownClass(cls):
     @classmethod
     def setUpClass(cls):
         cls._exit_stack = contextlib.ExitStack()
-        cls._exit_stack.enter_context(patch.object(config, "log_level", logging.DEBUG))
         cls._exit_stack.enter_context(
             patch.object(config, "raise_on_backend_error", True)
         )

torchdynamo/utils.py

@@ -8,6 +8,7 @@
 import gc
 import inspect
 import itertools
+import logging
 import logging.config
 import math
 import operator
@@ -29,9 +30,8 @@
 from torch import fx
 from torch.nn.modules.lazy import LazyModuleMixin
 
-import torchdynamo.config
-
-from . import config
+import torchdynamo
+import torchdynamo.config as config
 
 counters = collections.defaultdict(collections.Counter)
 troubleshooting_url = (
@@ -117,6 +117,20 @@ def fmt_fn(values, item_fn=lambda x: x):
         return headers, values
 
 
+# Return all loggers that torchdynamo is responsible for
+def get_loggers():
+    return [
+        logging.getLogger("torchdynamo"),
+        logging.getLogger("torchinductor"),
+    ]
+
+
+# Set the level of all loggers that torchdynamo is responsible for
+def set_loggers_level(level):
+    for logger in get_loggers():
+        logger.setLevel(level)
+
+
 LOGGING_CONFIG = {
     "version": 1,
     "formatters": {
@@ -146,18 +160,17 @@ def fmt_fn(values, item_fn=lambda x: x):
 }
 
 
+# initialize torchdynamo loggers
 def init_logging():
     if "PYTEST_CURRENT_TEST" not in os.environ:
         logging.config.dictConfig(LOGGING_CONFIG)
-        td_logger = logging.getLogger("torchdynamo")
-        td_logger.setLevel(config.log_level)
-        ti_logger = logging.getLogger("torchinductor")
-        ti_logger.setLevel(config.log_level)
         if config.log_file_name is not None:
             log_file = logging.FileHandler(config.log_file_name)
             log_file.setLevel(config.log_level)
-            td_logger.addHandler(log_file)
-            ti_logger.addHandler(log_file)
+            for logger in get_loggers():
+                logger.addHandler(log_file)
+
+    set_loggers_level(config.log_level)
 
 
 # filter out all frames after entering dynamo
@@ -793,14 +806,14 @@ def format_func_info(code):
 
 @contextlib.contextmanager
 def disable_cache_limit():
-    prior = torchdynamo.config.cache_size_limit
-    torchdynamo.config.cache_size_limit = sys.maxsize
+    prior = config.cache_size_limit
+    config.cache_size_limit = sys.maxsize
 
     try:
         yield
     finally:
         pass
-        torchdynamo.config.cache_size_limit = prior
+        config.cache_size_limit = prior
 
 
 # map from transformed code back to original user code