ahmedcs / flperf Goto Github PK

This repository contains an ongoing benchmarking project for Federated Learning (FL). We will release a standardized benchmark with efficient simulator and real-world datasets pretty soon! Please stay tuned! This benchmark is used in OSDI '21 paper "Efficient Federated Learning via Guided Participant Selection" (preprint here).

To download the dataset, run (or download from here manually):

git clone https://github.com/SymbioticLab/FLPerf.git
cd FLPerf
# Download all datasets from DropBox. Check download.sh -h for more details
bash download.sh -A  # If fail, bash download_google.sh from Google Drive

Note that no details were kept of any of the participants age, gender, or location, and random ids were assigned to each individual.

We provide real-world datasets for the federated learning community, and plan to release much more soon! Each is associated with its training and testing dataset. A summary of statistics for training datasets can be found in Table, and you can refer to each folder for more details.

A speech recognition dataset with over ten thousand clips of one-second-long duration. Each clip contains one of the 35 common words (e.g., digits zero to nine, "Yes", "No", "Up", "Down") spoken by thousands of different people.

OpenImage is a vision dataset collected from Flickr, an image and video hosting service. It contains a total of 16M bounding boxes for 600 object classes (e.g., Microwave oven). We clean up the dataset according to the provided indices of clients.

Reddit (StackOverflow) consists of comments from the Reddit (StackOverflow) website. It has been widely used for language modeling tasks, and we consider each user as a client. In our benchmark, we restrict to the 30k most frequently used words, and represent each sentence as a sequence of indices corresponding to these 30k frequently used words. We use Transformers to tokenize these sequences with a block size 64.

We use the AIBench dataset and MobiPerf dataset. AIBench dataset provides the computation capacity of different models across a wide range of devices. As specified in real FL deployments, we focus on the capability of mobile devices that have > 2GB RAM in this benchmark. To understand the network capacity of these devices, we clean up the MobiPerf dataset, and provide the available bandwidth when they are connected with WiFi, which is preferred in FL as well.

We use a large-scale real-world user behavior dataset from FLASH. It comes from a popular input method app (IMA) that can be downloaded from Google Play, and covers 136k users and spans one week from January 31st to February 6th in 2020. This dataset includes 180 million trace items (e.g., battery charge or screen lock) and we consider user devices that are in charging to be available, as specified in real FL deployments.

We plan to release a simulator for standardized benchmarking that can support the simulation of millions of clients soon.

please consider to cite our paper if you use the code or data in your research project.

@inproceedings{kuiper-osdi21,
  title={Kuiper: Efficient Federated Learning via Guided Participant Selection},
  author={Fan Lai, Xiangfeng Zhu, Harsha V. Madhyastha, Mosharaf Chowdhury},
  booktitle={USENIX Symposium on Operating Systems Design and Implementation (OSDI)},
  year={2021}
}

Fan Lai ([email protected]), Yinwei Dai ([email protected]) and Xiangfeng Zhu ([email protected]).