Merge branch 'main' into fix-aggregate-inplace

.editorconfig

@@ -16,6 +16,14 @@ profile = black
 indent_style = space
 indent_size = 2
 
+[*.md]
+indent_style = space
+indent_size = 2
+
 [*.yml]
 indent_style = space
 indent_size = 2
+
+[*.toml]
+indent_style = space
+indent_size = 4

.github/workflows/datasets-e2e.yml

@@ -45,7 +45,7 @@ jobs:
       - name: Bootstrap
         uses: ./.github/actions/bootstrap
         with:
-          python-version: 3.8
+          python-version: 3.9
       - name: Install dependencies
         run: python -m poetry install
       - name: Run tests

.github/workflows/datasets.yml

@@ -37,7 +37,7 @@ jobs:
         # In case of a mismatch, the job has to download Python to install it.
         # Note: Due to a bug in actions/setup-python, we have to put "3.10" in
         # quotes as it will otherwise assume "3.1"
-        python: [3.8, 3.9, '3.10', '3.11']
+        python: ['3.9', '3.10', '3.11']
 
     name: Python ${{ matrix.python }}
 

benchmarks/flowertune-llm/README.md

@@ -1,4 +1,4 @@
-![](_static/flower_llm.png)
+[![FlowerTune LLM Leaderboard](_static/flower_llm.png)](https://flower.ai/benchmarks/llm-leaderboard)
 
 # FlowerTune LLM Leaderboard
 
@@ -27,15 +27,16 @@ flwr new --framework=FlowerTune
 The `flwr new` command will generate a directory with the following structure:
 
 ```bash
-<project-name>
-       ├── README.md           # <- Instructions
-       ├── pyproject.toml      # <- Environment dependencies and configs
-       └── <project_name>
-                  ├── client_app.py   # <- Flower ClientApp build
-                  ├── dataset.py      # <- Dataset and tokenizer build
-                  ├── models.py       # <- Model build
-                  ├── server_app.py   # <- Flower ServerApp build
-                  └── strategy.py     # <- Flower strategy build
+<project_name>
+├── README.md           # Instructions
+├── pyproject.toml      # Environment dependencies and configs
+└── <project_name>
+    ├── __init__.py
+    ├── client_app.py   # Flower ClientApp build
+    ├── dataset.py      # Dataset and tokenizer build
+    ├── models.py       # Model build
+    ├── server_app.py   # Flower ServerApp build
+    └── strategy.py     # Flower strategy build
 ```
 
 This can serve as the starting point for you to build up your own federated LLM fine-tuning methods.

benchmarks/flowertune-llm/evaluation/README.md

@@ -5,7 +5,7 @@ If you are participating [LLM Leaderboard](https://flower.ai/benchmarks/llm-lead
 
 ## How to run
 
-Navigate to the directory corresponding to your selected challenge (`general NLP`, `finance`, `medical`, or `code`) and follow the instructions there to execute the evaluation.
+Navigate to the directory corresponding to your selected challenge ([`general NLP`](https://github.com/adap/flower/tree/main/benchmarks/flowertune-llm/evaluation/general-nlp), [`finance`](https://github.com/adap/flower/tree/main/benchmarks/flowertune-llm/evaluation/finance), [`medical`](https://github.com/adap/flower/tree/main/benchmarks/flowertune-llm/evaluation/medical), or [`code`](https://github.com/adap/flower/tree/main/benchmarks/flowertune-llm/evaluation/code)) and follow the instructions there to execute the evaluation.
 
 > [!NOTE]  
 > If you wish to participate in the LLM Leaderboard, you must not modify the evaluation code and should use the exact command provided in the respective directory to run the evaluation.

datasets/flwr_datasets/common/telemetry.py

@@ -25,7 +25,7 @@
 from concurrent.futures import Future, ThreadPoolExecutor
 from enum import Enum, auto
 from pathlib import Path
-from typing import Any, Dict, List, Optional, Union, cast
+from typing import Any, Optional, Union, cast
 
 from flwr_datasets.common.version import package_name, package_version
 
@@ -114,7 +114,7 @@ class EventType(str, Enum):
     # The type signature is not compatible with mypy, pylint and flake8
     # so each of those needs to be disabled for this line.
     # pylint: disable-next=no-self-argument,arguments-differ,line-too-long
-    def _generate_next_value_(name: str, start: int, count: int, last_values: List[Any]) -> Any:  # type: ignore # noqa: E501
+    def _generate_next_value_(name: str, start: int, count: int, last_values: list[Any]) -> Any:  # type: ignore # noqa: E501
         return name
 
     PING = auto()
@@ -127,7 +127,7 @@ def _generate_next_value_(name: str, start: int, count: int, last_values: List[A
 
 # Use the ThreadPoolExecutor with max_workers=1 to have a queue
 # and also ensure that telemetry calls are not blocking.
-state: Dict[str, Union[Optional[str], Optional[ThreadPoolExecutor]]] = {
+state: dict[str, Union[Optional[str], Optional[ThreadPoolExecutor]]] = {
     # Will be assigned ThreadPoolExecutor(max_workers=1)
     # in event() the first time it's required
     "executor": None,
@@ -143,7 +143,7 @@ def _generate_next_value_(name: str, start: int, count: int, last_values: List[A
 # pylint: disable-next=unsubscriptable-object
 def event(
     event_type: EventType,
-    event_details: Optional[Dict[str, Any]] = None,
+    event_details: Optional[dict[str, Any]] = None,
 ) -> Future:  # type: ignore
     """Submit create_event to ThreadPoolExecutor to avoid blocking."""
     if state["executor"] is None:
@@ -155,7 +155,7 @@ def event(
     return result
 
 
-def create_event(event_type: EventType, event_details: Optional[Dict[str, Any]]) -> str:
+def create_event(event_type: EventType, event_details: Optional[dict[str, Any]]) -> str:
     """Create telemetry event."""
     if state["source"] is None:
         state["source"] = _get_source_id()

datasets/flwr_datasets/common/typing.py

@@ -15,12 +15,12 @@
 """Flower Datasets type definitions."""
 
 
-from typing import Any, List
+from typing import Any
 
 import numpy as np
 import numpy.typing as npt
 
 NDArray = npt.NDArray[Any]
 NDArrayInt = npt.NDArray[np.int_]
 NDArrayFloat = npt.NDArray[np.float_]
-NDArrays = List[NDArray]
+NDArrays = list[NDArray]

datasets/flwr_datasets/common/version.py

@@ -19,15 +19,14 @@
 
 
 import importlib.metadata as importlib_metadata
-from typing import Tuple
 
 
-def _check_package(name: str) -> Tuple[str, str]:
+def _check_package(name: str) -> tuple[str, str]:
     version: str = importlib_metadata.version(name)
     return name, version
 
 
-def _version() -> Tuple[str, str]:
+def _version() -> tuple[str, str]:
     """Read and return Flower Dataset package name and version.
 
     Returns

datasets/flwr_datasets/federated_dataset_test.py

@@ -17,7 +17,7 @@
 
 
 import unittest
-from typing import Dict, Union
+from typing import Union
 from unittest.mock import Mock, patch
 
 import numpy as np
@@ -385,7 +385,7 @@ def test_dict_of_partitioners_passes_partitioners(self) -> None:
         """Test if partitioners are passed directly (no recreation)."""
         num_train_partitions = 100
         num_test_partitions = 100
-        partitioners: Dict[str, Union[Partitioner, int]] = {
+        partitioners: dict[str, Union[Partitioner, int]] = {
             "train": IidPartitioner(num_partitions=num_train_partitions),
             "test": IidPartitioner(num_partitions=num_test_partitions),
         }
@@ -419,7 +419,7 @@ def test_mixed_type_partitioners_passes_instantiated_partitioners(self) -> None:
         """Test if an instantiated partitioner is passed directly."""
         num_train_partitions = 100
         num_test_partitions = 100
-        partitioners: Dict[str, Union[Partitioner, int]] = {
+        partitioners: dict[str, Union[Partitioner, int]] = {
             "train": IidPartitioner(num_partitions=num_train_partitions),
             "test": num_test_partitions,
         }
@@ -433,7 +433,7 @@ def test_mixed_type_partitioners_creates_from_int(self) -> None:
         """Test if an IidPartitioner partitioner is created."""
         num_train_partitions = 100
         num_test_partitions = 100
-        partitioners: Dict[str, Union[Partitioner, int]] = {
+        partitioners: dict[str, Union[Partitioner, int]] = {
             "train": IidPartitioner(num_partitions=num_train_partitions),
             "test": num_test_partitions,
         }

datasets/flwr_datasets/metrics/utils.py

@@ -16,7 +16,7 @@
 
 
 import warnings
-from typing import List, Optional, Union
+from typing import Optional, Union
 
 import pandas as pd
 
@@ -206,7 +206,7 @@ def compute_frequencies(
 
 
 def _compute_counts(
-    labels: Union[List[int], List[str]], unique_labels: Union[List[int], List[str]]
+    labels: Union[list[int], list[str]], unique_labels: Union[list[int], list[str]]
 ) -> pd.Series:
     """Compute the count of labels when taking into account all possible labels.
 
@@ -237,7 +237,7 @@ def _compute_counts(
 
 
 def _compute_frequencies(
-    labels: Union[List[int], List[str]], unique_labels: Union[List[int], List[str]]
+    labels: Union[list[int], list[str]], unique_labels: Union[list[int], list[str]]
 ) -> pd.Series:
     """Compute the distribution of labels when taking into account all possible labels.
 

datasets/flwr_datasets/mock_utils_test.py

@@ -19,7 +19,7 @@
 import random
 import string
 from datetime import datetime, timedelta
-from typing import Any, Dict, List, Optional, Set, Tuple, Union
+from typing import Any, Optional, Union
 
 import numpy as np
 from PIL import Image
@@ -30,7 +30,7 @@
 
 def _generate_artificial_strings(
     num_rows: int, num_unique: int, string_length: int, seed: int = 42
-) -> List[str]:
+) -> list[str]:
     """Create list of strings for categories or labels mocking.
 
     Note to keep the seed the same if you reuse this function for in creation of the
@@ -53,7 +53,7 @@ def _generate_artificial_strings(
         List of generated strings.
     """
     random.seed(seed)
-    unique_strings: Set[str] = set()
+    unique_strings: set[str] = set()
     while len(unique_strings) < num_unique:
         random_str = "".join(
             random.choices(string.ascii_letters + string.digits, k=string_length)
@@ -68,7 +68,7 @@ def _generate_artificial_strings(
     return artificial_column
 
 
-def _generate_artificial_categories(num_rows: int, choices: List[Any]) -> List[str]:
+def _generate_artificial_categories(num_rows: int, choices: list[Any]) -> list[str]:
     """Create list of strings from given `choices` list."""
     artificial_column = choices.copy()
     remaining_to_allocate = num_rows - len(choices)
@@ -82,7 +82,7 @@ def _generate_random_word(length: int) -> str:
     return "".join(random.choices(string.ascii_letters, k=length))
 
 
-def _generate_random_text_column(num_rows: int, length: int) -> List[str]:
+def _generate_random_text_column(num_rows: int, length: int) -> list[str]:
     """Generate a list of random text of specified length."""
     text_col = []
     for _ in range(num_rows):
@@ -98,7 +98,7 @@ def _generate_random_sentence(
 ) -> str:
     """Generate a random sentence with words of random lengths."""
     sentence_length = random.randint(min_sentence_length, max_sentence_length)
-    sentence: List[str] = []
+    sentence: list[str] = []
     while len(" ".join(sentence)) < sentence_length:
         word_length = random.randint(min_word_length, max_word_length)
         word = _generate_random_word(word_length)
@@ -112,7 +112,7 @@ def _generate_random_sentences(
     max_word_length: int,
     min_sentence_length: int,
     max_sentence_length: int,
-) -> List[str]:
+) -> list[str]:
     """Generate a list of random sentences."""
     text_col = [
         _generate_random_sentence(
@@ -123,7 +123,7 @@ def _generate_random_sentences(
     return text_col
 
 
-def _make_num_rows_none(column: List[Any], num_none: int) -> List[Any]:
+def _make_num_rows_none(column: list[Any], num_none: int) -> list[Any]:
     """Assign none num_none times to the given list."""
     column_copy = column.copy()
     none_positions = random.sample(range(len(column_copy)), num_none)
@@ -154,29 +154,29 @@ def _generate_random_date_column(
     end_date: datetime,
     date_format: str = "%a %b %d %H:%M:%S %Y",
     as_string: bool = True,
-) -> List[Union[str, datetime]]:
+) -> list[Union[str, datetime]]:
     """Generate a list of random dates."""
     return [
         _generate_random_date(start_date, end_date, date_format, as_string)
         for _ in range(num_rows)
     ]
 
 
-def _generate_random_int_column(num_rows: int, min_int: int, max_int: int) -> List[int]:
+def _generate_random_int_column(num_rows: int, min_int: int, max_int: int) -> list[int]:
     """Generate a list of ints."""
     return [random.randint(min_int, max_int) for _ in range(num_rows)]
 
 
-def _generate_random_bool_column(num_rows: int) -> List[bool]:
+def _generate_random_bool_column(num_rows: int) -> list[bool]:
     """Generate a list of bools."""
     return [random.choice([True, False]) for _ in range(num_rows)]
 
 
 def _generate_random_image_column(
     num_rows: int,
-    image_size: Union[Tuple[int, int], Tuple[int, int, int]],
+    image_size: Union[tuple[int, int], tuple[int, int, int]],
     simulate_type: str,
-) -> List[Any]:
+) -> list[Any]:
     """Simulate the images with the format that is found in HF Hub.
 
     Directly using `Image.fromarray` does not work because it creates `PIL.Image.Image`.
@@ -207,7 +207,7 @@ def generate_random_audio_column(
     num_rows: int,
     sampling_rate: int,
     length_in_samples: int,
-) -> List[Dict[str, Any]]:
+) -> list[dict[str, Any]]:
     """Simulate the audio column.
 
     Audio column in the datset is comprised from an array or floats, sample_rate and a
@@ -365,8 +365,8 @@ def _mock_speach_commands(num_rows: int) -> Dataset:
 
 def _load_mocked_dataset(
     dataset_name: str,
-    num_rows: List[int],
-    split_names: List[str],
+    num_rows: list[int],
+    split_names: list[str],
     subset: str = "",
 ) -> DatasetDict:
     dataset_dict = {}
@@ -380,7 +380,7 @@ def _load_mocked_dataset(
 def _load_mocked_dataset_by_partial_download(
     dataset_name: str,
     split_name: str,
-    skip_take_list: List[Tuple[int, int]],
+    skip_take_list: list[tuple[int, int]],
     subset_name: Optional[str] = None,
 ) -> Dataset:
     """Download a partial dataset.
@@ -423,8 +423,8 @@ def _load_mocked_dataset_by_partial_download(
 
 def _load_mocked_dataset_dict_by_partial_download(
     dataset_name: str,
-    split_names: List[str],
-    skip_take_lists: List[List[Tuple[int, int]]],
+    split_names: list[str],
+    skip_take_lists: list[list[tuple[int, int]]],
     subset_name: Optional[str] = None,
 ) -> DatasetDict:
     """Like _load_mocked_dataset_by_partial_download but for many splits."""

datasets/flwr_datasets/partitioner/__init__.py

@@ -27,6 +27,7 @@
 from .partitioner import Partitioner
 from .pathological_partitioner import PathologicalPartitioner
 from .shard_partitioner import ShardPartitioner
+from .size_partitioner import SizePartitioner
 from .square_partitioner import SquarePartitioner
 
 __all__ = [
@@ -42,5 +43,6 @@
     "Partitioner",
     "PathologicalPartitioner",
     "ShardPartitioner",
+    "SizePartitioner",
     "SquarePartitioner",
 ]