Merge branch 'main' into fds-add-typing-module

.github/workflows/_docker-build.yml

@@ -98,7 +98,7 @@ jobs:
           touch "/tmp/digests/${digest#sha256:}"
 
       - name: Upload digest
-        uses: actions/upload-artifact@c7d193f32edcb7bfad88892161225aeda64e9392 # v4.0.0
+        uses: actions/upload-artifact@1eb3cb2b3e0f29609092a73eb033bb759a334595 # v4.1.0
         with:
           name: digests-${{ steps.build-id.outputs.id }}-${{ matrix.platform.name }}
           path: /tmp/digests/*
@@ -114,7 +114,7 @@ jobs:
       metadata: ${{ steps.meta.outputs.json }}
     steps:
       - name: Download digests
-        uses: actions/download-artifact@f44cd7b40bfd40b6aa1cc1b9b5b7bf03d3c67110 # v4.1.0
+        uses: actions/download-artifact@6b208ae046db98c579e8a3aa621ab581ff575935 # v4.1.1
         with:
           pattern: digests-${{ needs.build.outputs.build-id }}-*
           path: /tmp/digests

.github/workflows/swift.yml

@@ -20,7 +20,7 @@ jobs:
     name: Test
     runs-on: macos-latest
     steps:
-      - uses: fwal/setup-swift@f51889efb55dccf13be0ee727e3d6c89a096fb4c
+      - uses: fwal/setup-swift@cdbe0f7f4c77929b6580e71983e8606e55ffe7e4
         with:
           swift-version: 5
       - uses: actions/checkout@v4
@@ -31,7 +31,7 @@ jobs:
     runs-on: macos-latest
     name: Build docs
     steps:
-      - uses: fwal/setup-swift@f51889efb55dccf13be0ee727e3d6c89a096fb4c
+      - uses: fwal/setup-swift@cdbe0f7f4c77929b6580e71983e8606e55ffe7e4
         with:
           swift-version: 5
       - uses: actions/checkout@v4
@@ -44,7 +44,7 @@ jobs:
     runs-on: macos-latest
     name: Deploy docs 
     steps:
-      - uses: fwal/setup-swift@f51889efb55dccf13be0ee727e3d6c89a096fb4c
+      - uses: fwal/setup-swift@cdbe0f7f4c77929b6580e71983e8606e55ffe7e4
         with:
           swift-version: 5
       - uses: actions/checkout@v4

doc/source/how-to-install-flower.rst

@@ -11,6 +11,9 @@ Flower requires at least `Python 3.8 <https://docs.python.org/3.8/>`_, but `Pyth
 Install stable release
 ----------------------
 
+Using pip
+~~~~~~~~~
+
 Stable releases are available on `PyPI <https://pypi.org/project/flwr/>`_::
 
   python -m pip install flwr
@@ -20,6 +23,25 @@ For simulations that use the Virtual Client Engine, ``flwr`` should be installed
   python -m pip install flwr[simulation]
 
 
+Using conda (or mamba)
+~~~~~~~~~~~~~~~~~~~~~~
+
+Flower can also be installed from the ``conda-forge`` channel.
+
+If you have not added ``conda-forge`` to your channels, you will first need to run the following::
+
+  conda config --add channels conda-forge
+  conda config --set channel_priority strict
+
+Once the ``conda-forge`` channel has been enabled, ``flwr`` can be installed with ``conda``::
+
+  conda install flwr
+
+or with ``mamba``::
+
+  mamba install flwr
+
+
 Verify installation
 -------------------
 

examples/advanced-pytorch/README.md

@@ -1,6 +1,6 @@
 # Advanced Flower Example (PyTorch)
 
-This example demonstrates an advanced federated learning setup using Flower with PyTorch. It differs from the quickstart example in the following ways:
+This example demonstrates an advanced federated learning setup using Flower with PyTorch. This example uses [Flower Datasets](https://flower.dev/docs/datasets/) and it differs from the quickstart example in the following ways:
 
 - 10 clients (instead of just 2)
 - Each client holds a local dataset of 5000 training examples and 1000 test examples (note that using the `run.sh` script will only select 10 data samples by default, as the `--toy` argument is set).
@@ -59,12 +59,13 @@ pip install -r requirements.txt
 
 The included `run.sh` will start the Flower server (using `server.py`),
 sleep for 2 seconds to ensure that the server is up, and then start 10 Flower clients (using `client.py`) with only a small subset of the data (in order to run on any machine),
-but this can be changed by removing the `--toy True` argument in the script. You can simply start everything in a terminal as follows:
+but this can be changed by removing the `--toy` argument in the script. You can simply start everything in a terminal as follows:
 
 ```shell
-poetry run ./run.sh
+# After activating your environment
+./run.sh
 ```
 
 The `run.sh` script starts processes in the background so that you don't have to open eleven terminal windows. If you experiment with the code example and something goes wrong, simply using `CTRL + C` on Linux (or `CMD + C` on macOS) wouldn't normally kill all these processes, which is why the script ends with `trap "trap - SIGTERM && kill -- -$$" SIGINT SIGTERM EXIT` and `wait`. This simply allows you to stop the experiment using `CTRL + C` (or `CMD + C`). If you change the script and anything goes wrong you can still use `killall python` (or `killall python3`) to kill all background processes (or a more specific command if you have other Python processes running that you don't want to kill).
 
-You can also manually run `poetry run python3 server.py` and `poetry run python3 client.py` for as many clients as you want but you have to make sure that each command is ran in a different terminal window (or a different computer on the network).
+You can also manually run `python3 server.py` and `python3 client.py --client-id <ID>` for as many clients as you want but you have to make sure that each command is run in a different terminal window (or a different computer on the network).

examples/advanced-pytorch/client.py

@@ -6,16 +6,17 @@
 import argparse
 from collections import OrderedDict
 import warnings
+import datasets
 
 warnings.filterwarnings("ignore")
 
 
 class CifarClient(fl.client.NumPyClient):
     def __init__(
         self,
-        trainset: torchvision.datasets,
-        testset: torchvision.datasets,
-        device: str,
+        trainset: datasets.Dataset,
+        testset: datasets.Dataset,
+        device: torch.device,
         validation_split: int = 0.1,
     ):
         self.device = device
@@ -41,17 +42,14 @@ def fit(self, parameters, config):
         batch_size: int = config["batch_size"]
         epochs: int = config["local_epochs"]
 
-        n_valset = int(len(self.trainset) * self.validation_split)
+        train_valid = self.trainset.train_test_split(self.validation_split)
+        trainset = train_valid["train"]
+        valset = train_valid["test"]
 
-        valset = torch.utils.data.Subset(self.trainset, range(0, n_valset))
-        trainset = torch.utils.data.Subset(
-            self.trainset, range(n_valset, len(self.trainset))
-        )
+        train_loader = DataLoader(trainset, batch_size=batch_size, shuffle=True)
+        val_loader = DataLoader(valset, batch_size=batch_size)
 
-        trainLoader = DataLoader(trainset, batch_size=batch_size, shuffle=True)
-        valLoader = DataLoader(valset, batch_size=batch_size)
-
-        results = utils.train(model, trainLoader, valLoader, epochs, self.device)
+        results = utils.train(model, train_loader, val_loader, epochs, self.device)
 
         parameters_prime = utils.get_model_params(model)
         num_examples_train = len(trainset)
@@ -73,13 +71,13 @@ def evaluate(self, parameters, config):
         return float(loss), len(self.testset), {"accuracy": float(accuracy)}
 
 
-def client_dry_run(device: str = "cpu"):
+def client_dry_run(device: torch.device = "cpu"):
     """Weak tests to check whether all client methods are working as expected."""
 
     model = utils.load_efficientnet(classes=10)
     trainset, testset = utils.load_partition(0)
-    trainset = torch.utils.data.Subset(trainset, range(10))
-    testset = torch.utils.data.Subset(testset, range(10))
+    trainset = trainset.select(range(10))
+    testset = testset.select(range(10))
     client = CifarClient(trainset, testset, device)
     client.fit(
         utils.get_model_params(model),
@@ -102,7 +100,7 @@ def main() -> None:
         help="Do a dry-run to check the client",
     )
     parser.add_argument(
-        "--partition",
+        "--client-id",
         type=int,
         default=0,
         choices=range(0, 10),
@@ -112,9 +110,7 @@ def main() -> None:
     )
     parser.add_argument(
         "--toy",
-        type=bool,
-        default=False,
-        required=False,
+        action='store_true',
         help="Set to true to quicky run the client using only 10 datasamples. \
         Useful for testing purposes. Default: False",
     )
@@ -136,12 +132,11 @@ def main() -> None:
         client_dry_run(device)
     else:
         # Load a subset of CIFAR-10 to simulate the local data partition
-        trainset, testset = utils.load_partition(args.partition)
+        trainset, testset = utils.load_partition(args.client_id)
 
         if args.toy:
-            trainset = torch.utils.data.Subset(trainset, range(10))
-            testset = torch.utils.data.Subset(testset, range(10))
-
+            trainset = trainset.select(range(10))
+            testset = testset.select(range(10))
         # Start Flower client
         client = CifarClient(trainset, testset, device)
 

examples/advanced-pytorch/pyproject.toml

@@ -14,6 +14,7 @@ authors = [
 [tool.poetry.dependencies]
 python = ">=3.8,<3.11"
 flwr = ">=1.0,<2.0"
+flwr-datasets = { extras = ["vision"], version = ">=0.0.2,<1.0.0" }
 torch = "1.13.1"
 torchvision = "0.14.1"
 validators = "0.18.2"

examples/advanced-pytorch/requirements.txt

@@ -1,4 +1,5 @@
 flwr>=1.0, <2.0
+flwr-datasets[vision]>=0.0.2, <1.0.0
 torch==1.13.1
 torchvision==0.14.1
 validators==0.18.2

examples/advanced-pytorch/run.sh

@@ -2,20 +2,17 @@
 set -e
 cd "$( cd "$( dirname "${BASH_SOURCE[0]}" )" >/dev/null 2>&1 && pwd )"/
 
-# Download the CIFAR-10 dataset
-python -c "from torchvision.datasets import CIFAR10; CIFAR10('./dataset', download=True)"
-
 # Download the EfficientNetB0 model
 python -c "import torch; torch.hub.load( \
         'NVIDIA/DeepLearningExamples:torchhub', \
         'nvidia_efficientnet_b0', pretrained=True)"
 
-python server.py &
-sleep 3  # Sleep for 3s to give the server enough time to start
+python server.py --toy  &
+sleep 10  # Sleep for 10s to give the server enough time to start and dowload the dataset
 
 for i in `seq 0 9`; do
     echo "Starting client $i"
-    python client.py --partition=${i} --toy True &
+    python client.py --client-id=${i} --toy &
 done
 
 # Enable CTRL+C to stop all background processes

examples/advanced-pytorch/server.py

@@ -10,6 +10,8 @@
 
 import warnings
 
+from flwr_datasets import FederatedDataset
+
 warnings.filterwarnings("ignore")
 
 
@@ -39,18 +41,13 @@ def evaluate_config(server_round: int):
 def get_evaluate_fn(model: torch.nn.Module, toy: bool):
     """Return an evaluation function for server-side evaluation."""
 
-    # Load data and model here to avoid the overhead of doing it in `evaluate` itself
-    trainset, _, _ = utils.load_data()
-
-    n_train = len(trainset)
+    # Load data here to avoid the overhead of doing it in `evaluate` itself
+    centralized_data = utils.load_centralized_data()
     if toy:
         # use only 10 samples as validation set
-        valset = torch.utils.data.Subset(trainset, range(n_train - 10, n_train))
-    else:
-        # Use the last 5k training examples as a validation set
-        valset = torch.utils.data.Subset(trainset, range(n_train - 5000, n_train))
+        centralized_data = centralized_data.select(range(10))
 
-    valLoader = DataLoader(valset, batch_size=16)
+    val_loader = DataLoader(centralized_data, batch_size=16)
 
     # The `evaluate` function will be called after every round
     def evaluate(
@@ -63,7 +60,7 @@ def evaluate(
         state_dict = OrderedDict({k: torch.tensor(v) for k, v in params_dict})
         model.load_state_dict(state_dict, strict=True)
 
-        loss, accuracy = utils.test(model, valLoader)
+        loss, accuracy = utils.test(model, val_loader)
         return loss, {"accuracy": accuracy}
 
     return evaluate
@@ -79,9 +76,7 @@ def main():
     parser = argparse.ArgumentParser(description="Flower")
     parser.add_argument(
         "--toy",
-        type=bool,
-        default=False,
-        required=False,
+        action='store_true',
         help="Set to true to use only 10 datasamples for validation. \
             Useful for testing purposes. Default: False",
     )