lh-ycx / olaf-1

imc20

Leaf: A Benchmark for Federated Settings

What we have done to Leaf

Briefly speaking, we make Leaf simulate the federated learning process more in line with the fact.

Deadline

We add deadline setting for simulating failed upload and time out training. Now deadline follows a normal distribution in each round, and each client has the same deadline in one round. You can set the deadline's normal distribution parameters in the config file.

Device Type

Each client is bundled with a device type. Each device type has different training speeds and uploading time. We support the small/mid/big device, whose parameter can be set in the config file. We also support self-defined device type(-1) whose parameter you can set in the code manually for more complexed simulation. Note that if a client's device is not specified i.e. None, the program will use real training time instead of the simulation time, which is not recommended.

Round Failure

In federated settings, if there are not enough devices to upload the results in a round, then this round will be regarded as a failed round and the global model will not be updated. To simulate it, we add a update_frac parameter. If the uploaded fraction is smaller than update_frac, then this round will fail. You can also set it in the config file.

Config

To simplify the command line arguments, we move most of the parameters to a config file. Also, we add some other parameters as put above for better simulation. Here are some details.

# -1 for unlimited, >0 for round num
num_rounds -1
# learning rate
learning_rate 0.1
# evaluate the global model's loss and accuracy every 'eval_every' round
eval_every 3
# client num per round
clients_per_round 30
# batch size
batch_size 10
# random seed
seed 0
# run 'num_epochs' epochs in one round for each client
num_epochs 1
# deadline's μ and σ
round_ddl 30 5
# least update fraction for a successful round
update_frac 0.5
# μ and σ of big device's upload time
big_upload_time 5 1
# μ and σ of mid device's upload time
mid_upload_time 10 1
# μ and σ of small device's upload time
small_upload_time 15 1
# μ and σ of big device's training speed
big_speed 150 1
# μ and σ of mid device's training speed
mid_speed 100 1
# μ and σ of small device's training speed
small_speed 50 1

How to run it

git clone https://github.com/lh-ycx/leaf.git
pip3 install -r requirements.txt
cd leaf/data/shakespeare/
./preprocess.sh -s niid --sf 0.2 -k 0 -t sample -tf 0.8
cd ../../models/
python3 main.py

Resources

• Homepage: leaf.cmu.edu
• Paper: "LEAF: A Benchmark for Federated Settings"

Datasets

1. FEMNIST

• Overview: Image Dataset
• Details: 62 different classes (10 digits, 26 lowercase, 26 uppercase), images are 28 by 28 pixels (with option to make them all 128 by 128 pixels), 3500 users
• Task: Image Classification

2. Sentiment140

• Overview: Text Dataset of Tweets
• Details 660120 users
• Task: Sentiment Analysis

3. Shakespeare

• Overview: Text Dataset of Shakespeare Dialogues
• Details: 2288 users
• Task: Next-Character Prediction

4. Celeba

• Overview: Image Dataset based on the Large-scale CelebFaces Attributes Dataset
• Details: 9343 users (we exclude celebrities with less than 5 images)
• Task: Image Classification (Smiling vs. Not smiling)

5. Synthetic Dataset

• Overview: We propose a process to generate synthetic, challenging federated datasets. The high-level goal is to create devices whose true models are device-dependant. To see a description of the whole generative process, please refer to the paper
• Details: The user can customize the number of devices, the number of classes and the number of dimensions, among others
• Task: Classification

6. Reddit

• Overview: We preprocess the Reddit data released by pushshift.io corresponding to December 2017.
• Details: 1,660,820 users with a total of 56,587,343 comments.
• Task: Next-word Prediction.

Notes

• Install the libraries listed in requirements.txt
  • I.e. with pip: run pip3 install -r requirements.txt
• Go to directory of respective dataset for instructions on generating data
• models directory contains instructions on running baseline reference implementations