Add seed in train_test_split

datasets/doc/source/how-to-use-with-pytorch.rst

@@ -63,7 +63,7 @@ expected by a model with a convolutional layer.
 
 If you want to divide the dataset, you can use (at any point before passing the dataset to the DataLoader)::
 
-  partition_train_test = partition.train_test_split(test_size=0.2)
+  partition_train_test = partition.train_test_split(test_size=0.2, seed=42)
   partition_train = partition_train_test["train"]
   partition_test = partition_train_test["test"]
 

datasets/e2e/pytorch/pytorch_test.py

@@ -65,7 +65,7 @@ def _create_trainloader(self, batch_size: int) -> DataLoader:
         partition_id = 0
         fds = FederatedDataset(dataset=self.dataset_name, partitioners={"train": 100})
         partition = fds.load_partition(partition_id, "train")
-        partition_train_test = partition.train_test_split(test_size=0.2)
+        partition_train_test = partition.train_test_split(test_size=0.2, seed=42)
         partition_train_test = partition_train_test.map(
             lambda img: {"img": self.transforms(img)}, input_columns="img"
         )

datasets/e2e/scikit-learn/sklearn_test.py

@@ -29,7 +29,7 @@ def _get_partition_data(self):
         fds = FederatedDataset(dataset=self.dataset_name, partitioners={"train": 10})
         partition = fds.load_partition(partition_id, "train")
         partition.set_format("numpy")
-        partition_train_test = partition.train_test_split(test_size=0.2)
+        partition_train_test = partition.train_test_split(test_size=0.2, seed=42)
         X_train, y_train = partition_train_test["train"]["image"], partition_train_test[
             "train"]["label"]
         X_test, y_test = partition_train_test["test"]["image"], partition_train_test[

doc/source/tutorial-series-get-started-with-flower-pytorch.ipynb

@@ -145,7 +145,7 @@
     "    for partition_id in range(NUM_CLIENTS):\n",
     "        partition = fds.load_partition(partition_id, \"train\")\n",
     "        partition = partition.with_transform(apply_transforms)\n",
-    "        partition = partition.train_test_split(train_size=0.8)\n",
+    "        partition = partition.train_test_split(train_size=0.8, seed=42)\n",
     "        trainloaders.append(DataLoader(partition[\"train\"], batch_size=BATCH_SIZE))\n",
     "        valloaders.append(DataLoader(partition[\"test\"], batch_size=BATCH_SIZE))\n",
     "    testset = fds.load_split(\"test\").with_transform(apply_transforms)\n",

examples/advanced-pytorch/client.py

@@ -46,7 +46,7 @@ def fit(self, parameters, config):
         batch_size: int = config["batch_size"]
         epochs: int = config["local_epochs"]
 
-        train_valid = self.trainset.train_test_split(self.validation_split)
+        train_valid = self.trainset.train_test_split(self.validation_split, seed=42)
         trainset = train_valid["train"]
         valset = train_valid["test"]
 

examples/advanced-pytorch/utils.py

@@ -14,7 +14,7 @@ def load_partition(partition_id, toy: bool = False):
     fds = FederatedDataset(dataset="cifar10", partitioners={"train": 10})
     partition = fds.load_partition(partition_id)
     # Divide data on each node: 80% train, 20% test
-    partition_train_test = partition.train_test_split(test_size=0.2)
+    partition_train_test = partition.train_test_split(test_size=0.2, seed=42)
     partition_train_test = partition_train_test.with_transform(apply_transforms)
     return partition_train_test["train"], partition_train_test["test"]
 

examples/advanced-tensorflow/client.py

@@ -123,7 +123,7 @@ def load_partition(idx: int):
     partition.set_format("numpy")
 
     # Divide data on each node: 80% train, 20% test
-    partition = partition.train_test_split(test_size=0.2)
+    partition = partition.train_test_split(test_size=0.2, seed=42)
     x_train, y_train = partition["train"]["img"] / 255.0, partition["train"]["label"]
     x_test, y_test = partition["test"]["img"] / 255.0, partition["test"]["label"]
     return x_train, y_train, x_test, y_test

examples/app-pytorch/client.py

@@ -18,7 +18,6 @@
 
 # Define FlowerClient and client_fn
 class FlowerClient(NumPyClient):
-
     def fit(self, parameters, config):
         set_weights(net, parameters)
         results = train(net, trainloader, testloader, epochs=1, device=DEVICE)

examples/custom-mods/client.py

@@ -86,7 +86,6 @@ def wandb_mod(msg: Message, context: Context, app: ClientAppCallable) -> Message
 
         # if the `ClientApp` just processed a "fit" message, let's log some metrics to W&B
         if reply.metadata.message_type == MessageType.TRAIN and reply.has_content():
-
             metrics = reply.content.configs_records
 
             results_to_log = dict(metrics.get("fitres.metrics", ConfigsRecord()))

examples/embedded-devices/client_pytorch.py

@@ -108,7 +108,7 @@ def apply_transforms(batch):
     for partition_id in range(NUM_CLIENTS):
         partition = fds.load_partition(partition_id, "train")
         # Divide data on each node: 90% train, 10% test
-        partition = partition.train_test_split(test_size=0.1)
+        partition = partition.train_test_split(test_size=0.1, seed=42)
         partition = partition.with_transform(apply_transforms)
         trainsets.append(partition["train"])
         validsets.append(partition["test"])

examples/embedded-devices/client_tf.py

@@ -44,7 +44,7 @@ def prepare_dataset(use_mnist: bool):
         partition = fds.load_partition(partition_id, "train")
         partition.set_format("numpy")
         # Divide data on each node: 90% train, 10% test
-        partition = partition.train_test_split(test_size=0.1)
+        partition = partition.train_test_split(test_size=0.1, seed=42)
         x_train, y_train = (
             partition["train"][img_key] / 255.0,
             partition["train"]["label"],

examples/fl-dp-sa/fl_dp_sa/task.py

@@ -42,7 +42,7 @@ def load_data(partition_id):
     fds = FederatedDataset(dataset="mnist", partitioners={"train": 100})
     partition = fds.load_partition(partition_id)
     # Divide data on each node: 80% train, 20% test
-    partition_train_test = partition.train_test_split(test_size=0.2)
+    partition_train_test = partition.train_test_split(test_size=0.2, seed=42)
     pytorch_transforms = Compose([ToTensor(), Normalize((0.5,), (0.5,))])
 
     def apply_transforms(batch):

examples/flower-via-docker-compose/helpers/load_data.py

@@ -25,7 +25,7 @@ def load_data(data_sampling_percentage=0.5, client_id=1, total_clients=2):
     partition.set_format("numpy")
 
     # Divide data on each client: 80% train, 20% test
-    partition = partition.train_test_split(test_size=0.2)
+    partition = partition.train_test_split(test_size=0.2, seed=42)
     x_train, y_train = partition["train"]["img"] / 255.0, partition["train"]["label"]
     x_test, y_test = partition["test"]["img"] / 255.0, partition["test"]["label"]
 

examples/pytorch-from-centralized-to-federated/cifar.py

@@ -56,7 +56,7 @@ def load_data(partition_id: int):
     fds = FederatedDataset(dataset="cifar10", partitioners={"train": 10})
     partition = fds.load_partition(partition_id)
     # Divide data on each node: 80% train, 20% test
-    partition_train_test = partition.train_test_split(test_size=0.2)
+    partition_train_test = partition.train_test_split(test_size=0.2, seed=42)
     pytorch_transforms = Compose(
         [ToTensor(), Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))]
     )

examples/quickstart-huggingface/client.py

@@ -22,7 +22,7 @@ def load_data(partition_id):
     fds = FederatedDataset(dataset="imdb", partitioners={"train": 1_000})
     partition = fds.load_partition(partition_id)
     # Divide data: 80% train, 20% test
-    partition_train_test = partition.train_test_split(test_size=0.2)
+    partition_train_test = partition.train_test_split(test_size=0.2, seed=42)
 
     tokenizer = AutoTokenizer.from_pretrained(CHECKPOINT)
 

examples/quickstart-mlcube/dev/mnist.py

@@ -36,6 +36,7 @@ def create_directory(path: str) -> None:
 
 def download(task_args: List[str]) -> None:
     """Task: download.
+
     Input parameters:
         --data_dir
     """
@@ -81,6 +82,7 @@ def download(task_args: List[str]) -> None:
 
 def train(task_args: List[str]) -> None:
     """Task: train.
+
     Input parameters:
         --data_dir, --log_dir, --model_dir, --parameters_file
     """
@@ -175,6 +177,7 @@ def train(task_args: List[str]) -> None:
 
 def evaluate(task_args: List[str]) -> None:
     """Task: train.
+
     Input parameters:
         --data_dir, --log_dir, --model_dir, --parameters_file
     """

examples/quickstart-mlx/client.py

@@ -107,7 +107,7 @@ def evaluate(self, parameters, config):
 
     fds = FederatedDataset(dataset="mnist", partitioners={"train": 3})
     partition = fds.load_partition(partition_id=args.partition_id)
-    partition_splits = partition.train_test_split(test_size=0.2)
+    partition_splits = partition.train_test_split(test_size=0.2, seed=42)
 
     partition_splits["train"].set_format("numpy")
     partition_splits["test"].set_format("numpy")

examples/quickstart-pytorch-lightning/mnist.py

@@ -82,9 +82,11 @@ def load_data(partition):
 
     partition = partition.with_transform(apply_transforms)
     # 20 % for on federated evaluation
-    partition_full = partition.train_test_split(test_size=0.2)
+    partition_full = partition.train_test_split(test_size=0.2, seed=42)
     # 60 % for the federated train and 20 % for the federated validation (both in fit)
-    partition_train_valid = partition_full["train"].train_test_split(train_size=0.75)
+    partition_train_valid = partition_full["train"].train_test_split(
+        train_size=0.75, seed=42
+    )
     trainloader = DataLoader(
         partition_train_valid["train"],
         batch_size=32,

examples/quickstart-pytorch/client.py

@@ -74,7 +74,7 @@ def load_data(partition_id):
     fds = FederatedDataset(dataset="cifar10", partitioners={"train": 3})
     partition = fds.load_partition(partition_id)
     # Divide data on each node: 80% train, 20% test
-    partition_train_test = partition.train_test_split(test_size=0.2)
+    partition_train_test = partition.train_test_split(test_size=0.2, seed=42)
     pytorch_transforms = Compose(
         [ToTensor(), Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))]
     )

examples/quickstart-tensorflow/client.py

@@ -30,7 +30,7 @@
 partition.set_format("numpy")
 
 # Divide data on each node: 80% train, 20% test
-partition = partition.train_test_split(test_size=0.2)
+partition = partition.train_test_split(test_size=0.2, seed=42)
 x_train, y_train = partition["train"]["img"] / 255.0, partition["train"]["label"]
 x_test, y_test = partition["test"]["img"] / 255.0, partition["test"]["label"]
 

examples/simulation-pytorch/sim.ipynb

@@ -497,7 +497,7 @@
     "        client_dataset = dataset.load_partition(int(cid), \"train\")\n",
     "\n",
     "        # Now let's split it into train (90%) and validation (10%)\n",
-    "        client_dataset_splits = client_dataset.train_test_split(test_size=0.1)\n",
+    "        client_dataset_splits = client_dataset.train_test_split(test_size=0.1, seed=42)\n",
     "\n",
     "        trainset = client_dataset_splits[\"train\"]\n",
     "        valset = client_dataset_splits[\"test\"]\n",

examples/simulation-pytorch/sim.py

@@ -94,7 +94,7 @@ def client_fn(cid: str) -> fl.client.Client:
         client_dataset = dataset.load_partition(int(cid), "train")
 
         # Now let's split it into train (90%) and validation (10%)
-        client_dataset_splits = client_dataset.train_test_split(test_size=0.1)
+        client_dataset_splits = client_dataset.train_test_split(test_size=0.1, seed=42)
 
         trainset = client_dataset_splits["train"]
         valset = client_dataset_splits["test"]

examples/simulation-tensorflow/sim.ipynb

@@ -179,7 +179,7 @@
     "        client_dataset = dataset.load_partition(int(cid), \"train\")\n",
     "\n",
     "        # Now let's split it into train (90%) and validation (10%)\n",
-    "        client_dataset_splits = client_dataset.train_test_split(test_size=0.1)\n",
+    "        client_dataset_splits = client_dataset.train_test_split(test_size=0.1, seed=42)\n",
     "\n",
     "        trainset = client_dataset_splits[\"train\"].to_tf_dataset(\n",
     "            columns=\"image\", label_cols=\"label\", batch_size=32\n",

examples/simulation-tensorflow/sim.py

@@ -83,7 +83,7 @@ def client_fn(cid: str) -> fl.client.Client:
         client_dataset = dataset.load_partition(int(cid), "train")
 
         # Now let's split it into train (90%) and validation (10%)
-        client_dataset_splits = client_dataset.train_test_split(test_size=0.1)
+        client_dataset_splits = client_dataset.train_test_split(test_size=0.1, seed=42)
 
         trainset = client_dataset_splits["train"].to_tf_dataset(
             columns="image", label_cols="label", batch_size=32

examples/vit-finetune/client.py

@@ -8,9 +8,7 @@
 
 
 class FedViTClient(NumPyClient):
-
     def __init__(self, trainset):
-
         self.trainset = trainset
         self.model = get_model()
 

examples/vit-finetune/main.py

@@ -19,7 +19,6 @@
 
 
 def main():
-
     args = parser.parse_args()
 
     # To control the degree of parallelism

examples/whisper-federated-finetuning/utils.py

@@ -107,10 +107,10 @@ def prepare_silences_dataset(train_dataset, ratio_silence: float = 0.1) -> Datas
     """Generate silences for the train set.
 
     One of the classes in the SpeechCommands datatset is `silence`. However, the dataset
-    does not include clips of silence. It does however include 5 long files with different
-    background sounds. The taks of this function is to extract several (defined by `ratio_silence`)
-    one-second long clips from those background audio files. Later, those audio clips will be
-    included into the training set.
+    does not include clips of silence. It does however include 5 long files with
+    different background sounds. The taks of this function is to extract several
+    (defined by `ratio_silence`) one-second long clips from those background audio
+    files. Later, those audio clips will be included into the training set.
     """
     # retrieve original silence audio clips
     silences = [d for d in train_dataset if d["label"] == 35]
@@ -138,9 +138,9 @@ def prepare_silences_dataset(train_dataset, ratio_silence: float = 0.1) -> Datas
 def construct_client_mapping(full_trainset, num_clients: int = 100):
     """Create a mapping to partition the dataset into `num_client` buckets.
 
-    These buckets contain the same number of `spekaer_id` but likely different
-    number of training exampes since each `speaker_id` in SpeechCommands does
-    provide different amounts of data to the dataset.
+    These buckets contain the same number of `spekaer_id` but likely different number of
+    training exampes since each `speaker_id` in SpeechCommands does provide different
+    amounts of data to the dataset.
     """
     client_ids = list(set(full_trainset["speaker_id"]))
     client_ids.remove(
@@ -191,7 +191,7 @@ def set_params(model: torch.nn.ModuleList, params: List[fl.common.NDArrays]):
 
 
 def get_model(device, num_classes, compile: bool = True):
-    """Create model: Whisper-tiny Encoder + classification head"""
+    """Create model: Whisper-tiny Encoder + classification head."""
     encoder = WhisperForConditionalGeneration.from_pretrained(
         "openai/whisper-tiny"
     ).get_encoder()

src/py/flwr/cli/new/templates/app/code/task.pytorch.py.tpl

@@ -39,7 +39,7 @@ def load_data(partition_id, num_partitions):
     fds = FederatedDataset(dataset="cifar10", partitioners={"train": num_partitions})
     partition = fds.load_partition(partition_id)
     # Divide data on each node: 80% train, 20% test
-    partition_train_test = partition.train_test_split(test_size=0.2)
+    partition_train_test = partition.train_test_split(test_size=0.2, seed=42)
     pytorch_transforms = Compose(
         [ToTensor(), Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))]
     )