Add enable_autolog_hparams argument to Trainer. (#20593)

* Make hyperparam logging optional
* Modify docs
* Add to CHANGELOG.md
* Fix typos

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

Author: duydl

docs/source-pytorch/common/trainer.rst

@@ -1077,6 +1077,32 @@ With :func:`torch.inference_mode` disabled, you can enable the grad of your mode
     trainer = Trainer(inference_mode=False)
     trainer.validate(model)
 
+enable_autolog_hparams
+^^^^^^^^^^^^^^^^^^^^^^
+
+Whether to log hyperparameters at the start of a run. Defaults to True.
+
+.. testcode::
+
+    # default used by the Trainer
+    trainer = Trainer(enable_autolog_hparams=True)
+
+    # disable logging hyperparams
+    trainer = Trainer(enable_autolog_hparams=False)
+
+With the parameter set to false, you can add custom code to log hyperparameters.
+
+.. code-block:: python
+
+    model = LitModel()
+    trainer = Trainer(enable_autolog_hparams=False)
+    for logger in trainer.loggers:
+        if isinstance(logger, lightning.pytorch.loggers.CSVLogger):
+            logger.log_hyperparams(hparams_dict_1)
+        else:
+            logger.log_hyperparams(hparams_dict_2)
+
+You can also use `self.logger.log_hyperparams(...)` inside `LightningModule` to log.
 
 -----
 

src/lightning/pytorch/CHANGELOG.md

@@ -8,6 +8,9 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 
 ### Added
 
+- Add enable_autolog_hparams argument to Trainer ([#20593](https://github.com/Lightning-AI/pytorch-lightning/pull/20593))
+
+
 - Allow LightningCLI to use a customized argument parser class ([#20596](https://github.com/Lightning-AI/pytorch-lightning/pull/20596))
 
 

src/lightning/pytorch/trainer/trainer.py

@@ -128,6 +128,7 @@ def __init__(
         sync_batchnorm: bool = False,
         reload_dataloaders_every_n_epochs: int = 0,
         default_root_dir: Optional[_PATH] = None,
+        enable_autolog_hparams: bool = True,
     ) -> None:
         r"""Customize every aspect of training via flags.
 
@@ -290,6 +291,9 @@ def __init__(
                 Default: ``os.getcwd()``.
                 Can be remote file paths such as `s3://mybucket/path` or 'hdfs://path/'
 
+            enable_autolog_hparams: Whether to log hyperparameters at the start of a run.
+                Default: ``True``.
+
         Raises:
             TypeError:
                 If ``gradient_clip_val`` is not an int or float.
@@ -496,6 +500,8 @@ def __init__(
             num_sanity_val_steps,
         )
 
+        self.enable_autolog_hparams = enable_autolog_hparams
+
     def fit(
         self,
         model: "pl.LightningModule",
@@ -962,7 +968,9 @@ def _run(
             call._call_callback_hooks(self, "on_fit_start")
             call._call_lightning_module_hook(self, "on_fit_start")
 
-        _log_hyperparams(self)
+        # only log hparams if enabled
+        if self.enable_autolog_hparams:
+            _log_hyperparams(self)
 
         if self.strategy.restore_checkpoint_after_setup:
             log.debug(f"{self.__class__.__name__}: restoring module and callbacks from checkpoint path: {ckpt_path}")

tests/tests_pytorch/loggers/test_csv.py

@@ -163,7 +163,7 @@ def test_metrics_reset_after_save(tmp_path):
 
 
 @mock.patch(
-    # Mock the existance check, so we can simulate appending to the metrics file
+    # Mock the existence check, so we can simulate appending to the metrics file
     "lightning.fabric.loggers.csv_logs._ExperimentWriter._check_log_dir_exists"
 )
 def test_append_metrics_file(_, tmp_path):