Set the generator seed in `random_split` usages in the examples (#18651)

docs/source-pytorch/data/datamodule.rst

@@ -89,7 +89,9 @@ The equivalent DataModule just organizes the same exact code, but makes it reusa
             self.mnist_test = MNIST(self.data_dir, train=False)
             self.mnist_predict = MNIST(self.data_dir, train=False)
             mnist_full = MNIST(self.data_dir, train=True)
-            self.mnist_train, self.mnist_val = random_split(mnist_full, [55000, 5000])
+            self.mnist_train, self.mnist_val = random_split(
+                mnist_full, [55000, 5000], generator=torch.Generator().manual_seed(42)
+            )
 
         def train_dataloader(self):
             return DataLoader(self.mnist_train, batch_size=self.batch_size)
@@ -146,7 +148,9 @@ Here's a more realistic, complex DataModule that shows how much more reusable th
             # Assign train/val datasets for use in dataloaders
             if stage == "fit":
                 mnist_full = MNIST(self.data_dir, train=True, transform=self.transform)
-                self.mnist_train, self.mnist_val = random_split(mnist_full, [55000, 5000])
+                self.mnist_train, self.mnist_val = random_split(
+                    mnist_full, [55000, 5000], generator=torch.Generator().manual_seed(42)
+                )
 
             # Assign test dataset for use in dataloader(s)
             if stage == "test":
@@ -230,7 +234,9 @@ There are also data operations you might want to perform on every GPU. Use :meth
             # Assign Train/val split(s) for use in Dataloaders
             if stage == "fit":
                 mnist_full = MNIST(self.data_dir, train=True, download=True, transform=self.transform)
-                self.mnist_train, self.mnist_val = random_split(mnist_full, [55000, 5000])
+                self.mnist_train, self.mnist_val = random_split(
+                    mnist_full, [55000, 5000], generator=torch.Generator().manual_seed(42)
+                )
 
             # Assign Test split(s) for use in Dataloaders
             if stage == "test":

examples/fabric/language_model/train.py

@@ -63,7 +63,8 @@ def validate(fabric, model, val_dataloader):
 
 def get_dataloaders(dataset):
     n = len(dataset)
-    train_dataset, val_dataset, test_dataset = random_split(dataset, [n - 4000, 2000, 2000])
+    generator = torch.Generator().manual_seed(42)
+    train_dataset, val_dataset, test_dataset = random_split(dataset, [n - 4000, 2000, 2000], generator=generator)
     train_dataloader = DataLoader(train_dataset, batch_size=20, shuffle=True)
     val_dataloader = DataLoader(val_dataset, batch_size=20, shuffle=False)
     test_dataloader = DataLoader(test_dataset, batch_size=20, shuffle=False)

examples/pytorch/basics/autoencoder.py

@@ -156,7 +156,9 @@ class MyDataModule(LightningDataModule):
         super().__init__()
         dataset = MNIST(DATASETS_PATH, train=True, download=True, transform=transforms.ToTensor())
         self.mnist_test = MNIST(DATASETS_PATH, train=False, download=True, transform=transforms.ToTensor())
-        self.mnist_train, self.mnist_val = random_split(dataset, [55000, 5000])
+        self.mnist_train, self.mnist_val = random_split(
+            dataset, [55000, 5000], generator=torch.Generator().manual_seed(42)
+        )
         self.batch_size = batch_size
 
     def train_dataloader(self):

examples/pytorch/basics/backbone_image_classifier.py

@@ -105,7 +105,9 @@ class MyDataModule(LightningDataModule):
         super().__init__()
         dataset = MNIST(DATASETS_PATH, train=True, download=True, transform=transforms.ToTensor())
         self.mnist_test = MNIST(DATASETS_PATH, train=False, download=True, transform=transforms.ToTensor())
-        self.mnist_train, self.mnist_val = random_split(dataset, [55000, 5000])
+        self.mnist_train, self.mnist_val = random_split(
+            dataset, [55000, 5000], generator=torch.Generator().manual_seed(42)
+        )
         self.batch_size = batch_size
 
     def train_dataloader(self):

src/lightning/pytorch/core/datamodule.py

@@ -48,7 +48,9 @@ class LightningDataModule(DataHooks, HyperparametersMixin):
                 # make assignments here (val/train/test split)
                 # called on every process in DDP
                 dataset = RandomDataset(1, 100)
-                self.train, self.val, self.test = data.random_split(dataset, [80, 10, 10])
+                self.train, self.val, self.test = data.random_split(
+                    dataset, [80, 10, 10], generator=torch.Generator().manual_seed(42)
+                )
 
             def train_dataloader(self):
                 return data.DataLoader(self.train)

src/lightning/pytorch/demos/mnist_datamodule.py

@@ -200,7 +200,9 @@ class MNISTDataModule(LightningDataModule):
         dataset: Dataset = MNIST(self.data_dir, train=True, download=False, **extra)
         assert isinstance(dataset, Sized)
         train_length = len(dataset)
-        self.dataset_train, self.dataset_val = random_split(dataset, [train_length - self.val_split, self.val_split])
+        self.dataset_train, self.dataset_val = random_split(
+            dataset, [train_length - self.val_split, self.val_split], generator=torch.Generator().manual_seed(42)
+        )
 
     def train_dataloader(self) -> DataLoader:
         """MNIST train set removes a subset to use for validation."""