From 440db6a42fa67060b138338bf83c4f75a1652013 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Adrian=20W=C3=A4lchli?= <aedu.waelchli@gmail.com>
Date: Sat, 6 Mar 2021 03:41:28 +0100
Subject: [PATCH] bugfix

---
 pytorch_lightning/trainer/optimizers.py |  3 +++
 tests/trainer/test_trainer.py           | 32 ++++++++++++++++++++++++-
 2 files changed, 34 insertions(+), 1 deletion(-)

diff --git a/pytorch_lightning/trainer/optimizers.py b/pytorch_lightning/trainer/optimizers.py
index 5cafa438cffcc..a247fb92cd22f 100644
--- a/pytorch_lightning/trainer/optimizers.py
+++ b/pytorch_lightning/trainer/optimizers.py
@@ -27,7 +27,10 @@
 
 class TrainerOptimizersMixin(ABC):
 
+    _lightning_optimizers: Optional[List[LightningOptimizer]]
+
     def init_optimizers(self, model: LightningModule) -> Tuple[List, List, List]:
+        self._lightning_optimizers = None
         optim_conf = model.configure_optimizers()
         if optim_conf is None:
             rank_zero_warn(
diff --git a/tests/trainer/test_trainer.py b/tests/trainer/test_trainer.py
index 1cd979c863d37..632428e6f1874 100644
--- a/tests/trainer/test_trainer.py
+++ b/tests/trainer/test_trainer.py
@@ -19,7 +19,7 @@
 from copy import deepcopy
 from distutils.version import LooseVersion
 from pathlib import Path
-from unittest.mock import ANY, call, patch
+from unittest.mock import ANY, call, patch, Mock
 
 import cloudpickle
 import pytest
@@ -1785,3 +1785,33 @@ def backward(self, *args, **kwargs):
         "training_step",
         "backward",
     ]
+
+
+def test_init_optimizers_resets_lightning_optimizers(tmpdir):
+    """ Test that the Trainer resets the `lightning_optimizers` list everytime new optimizers get initialized. """
+
+    def compare_optimizers():
+        assert trainer.lightning_optimizers[0].optimizer is trainer.optimizers[0]
+
+    class OptimizerSpy(Callback):
+        def on_fit_start(self, *args, **kwargs):
+            compare_optimizers()
+
+    model = BoringModel()
+    model.lr = 0.2
+    trainer = Trainer(
+        default_root_dir=tmpdir,
+        max_epochs=1,
+        auto_lr_find=True,
+        callbacks=[OptimizerSpy()]
+    )
+
+    trainer.tune(model)
+    compare_optimizers()
+
+    trainer.fit(model)
+    compare_optimizers()
+
+    trainer.max_epochs = 2  # simulate multiple fit calls
+    trainer.fit(model)
+    compare_optimizers()