Fix disabled grads after call to predict (#6657)

CHANGELOG.md

@@ -191,6 +191,9 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Fixed comparing required versions ([#6434](https://github.com/PyTorchLightning/pytorch-lightning/pull/6434))
 
 
+- Fixed a bug where gradients were disabled after calling `Trainer.predict` ([#6657](https://github.com/PyTorchLightning/pytorch-lightning/pull/6657))
+
+
 ## [1.2.4] - 2021-03-16
 
 ### Changed

pytorch_lightning/trainer/trainer.py

@@ -800,6 +800,10 @@ class Trainer(
 
         results = self.predict_loop.on_predict_epoch_end()
         self.predict_loop.on_predict_end()
+
+        # re-enable grads
+        torch.set_grad_enabled(True)
+
         return results
 
     def run_sanity_check(self, ref_model):

tests/trainer/test_trainer.py

@@ -1450,6 +1450,19 @@ def test_trainer_predict_no_return(tmpdir):
         predict(tmpdir, None, None, 1, model=CustomBoringModel())
 
 
+def test_trainer_predict_grad(tmpdir):
+    class CustomBoringModel(BoringModel):
+
+        def predict_step(self, batch, batch_idx, dataloader_idx=None):
+            assert batch.expand_as(batch).grad_fn is None
+            return super().predict_step(batch, batch_idx, dataloader_idx)
+
+    predict(tmpdir, None, None, 1, model=CustomBoringModel())
+
+    x = torch.zeros(1, requires_grad=True)
+    assert x.expand_as(x).grad_fn is not None
+
+
 @pytest.mark.parametrize('datamodule', [False, True])
 def test_trainer_predict_cpu(tmpdir, datamodule):
     predict(tmpdir, None, None, 1, datamodule=datamodule)