Merge branch 'master' into feature/9580-rich-defaults

Author: bhimrazy

docs/source-pytorch/common/checkpointing_basic.rst

@@ -111,7 +111,7 @@ The LightningModule also has access to the Hyperparameters
 .. code-block:: python
 
     model = MyLightningModule.load_from_checkpoint("/path/to/checkpoint.ckpt")
-    print(model.learning_rate)
+    print(model.hparams.learning_rate)
 
 ----
 

examples/fabric/image_classifier/train_fabric.py

@@ -158,7 +158,7 @@ def run(hparams):
     # When using distributed training, use `fabric.save`
     # to ensure the current process is allowed to save a checkpoint
     if hparams.save_model:
-        fabric.save(model.state_dict(), "mnist_cnn.pt")
+        fabric.save(path="mnist_cnn.pt", state=model.state_dict())
 
 
 if __name__ == "__main__":

examples/fabric/kfold_cv/train_fabric.py

@@ -161,7 +161,7 @@ def run(hparams):
     # When using distributed training, use `fabric.save`
     # to ensure the current process is allowed to save a checkpoint
     if hparams.save_model:
-        fabric.save(model.state_dict(), "mnist_cnn.pt")
+        fabric.save(path="mnist_cnn.pt", state=model.state_dict())
 
 
 if __name__ == "__main__":

examples/fabric/tensor_parallel/train.py

@@ -67,7 +67,7 @@ def train():
     # See `fabric consolidate --help` if you need to convert the checkpoint to a single file
     fabric.print("Saving a (distributed) checkpoint ...")
     state = {"model": model, "optimizer": optimizer, "iteration": i}
-    fabric.save("checkpoint.pt", state)
+    fabric.save(path="checkpoint.pt", state=state)
 
     fabric.print("Training successfully completed!")
     fabric.print(f"Peak memory usage: {torch.cuda.max_memory_allocated() / 1e9:.02f} GB")

src/lightning/pytorch/CHANGELOG.md

@@ -28,6 +28,7 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 ### Fixed
 
 - fix progress bar console clearing for Rich `14.1+` ([#21016](https://github.com/Lightning-AI/pytorch-lightning/pull/21016))
+- fix `AdvancedProfiler` to handle nested profiling actions for Python 3.12+ ([#20809](https://github.com/Lightning-AI/pytorch-lightning/pull/20809))
 
 
 ---

src/lightning/pytorch/profilers/advanced.py

@@ -19,6 +19,7 @@
 import os
 import pstats
 import tempfile
+from collections import defaultdict
 from pathlib import Path
 from typing import Optional, Union
 
@@ -66,14 +67,15 @@ def __init__(
                 If you attempt to stop recording an action which was never started.
         """
         super().__init__(dirpath=dirpath, filename=filename)
-        self.profiled_actions: dict[str, cProfile.Profile] = {}
+        self.profiled_actions: dict[str, cProfile.Profile] = defaultdict(cProfile.Profile)
         self.line_count_restriction = line_count_restriction
         self.dump_stats = dump_stats
 
     @override
     def start(self, action_name: str) -> None:
-        if action_name not in self.profiled_actions:
-            self.profiled_actions[action_name] = cProfile.Profile()
+        # Disable all profilers before starting a new one
+        for pr in self.profiled_actions.values():
+            pr.disable()
         self.profiled_actions[action_name].enable()
 
     @override
@@ -114,7 +116,7 @@ def summary(self) -> str:
     @override
     def teardown(self, stage: Optional[str]) -> None:
         super().teardown(stage=stage)
-        self.profiled_actions = {}
+        self.profiled_actions.clear()
 
     def __reduce__(self) -> tuple:
         # avoids `TypeError: cannot pickle 'cProfile.Profile' object`

tests/tests_pytorch/profilers/test_profiler.py

@@ -336,6 +336,12 @@ def test_advanced_profiler_deepcopy(advanced_profiler):
     assert deepcopy(advanced_profiler)
 
 
+def test_advanced_profiler_nested(advanced_profiler):
+    """Ensure AdvancedProfiler does not raise ValueError for nested profiling actions (Python 3.12+ compatibility)."""
+    with advanced_profiler.profile("outer"), advanced_profiler.profile("inner"):
+        pass  # Should not raise ValueError
+
+
 @pytest.fixture
 def pytorch_profiler(tmp_path):
     return PyTorchProfiler(dirpath=tmp_path, filename="profiler")