FedPilot Config Generator
Create and customize your federated learning configuration
📋
Quick Start Templates
🚀
Quick Start - FMNIST CNN
cnn / fmnist
🚀
Non-IID CIFAR-10
resnet / cifar10
🚀
Privacy-Preserving FL
cnn / fmnist
🚀
Label-100 Decentralized
cnn / fmnist
⚙️
General Settings
Federation Id
*
Unique identifier for this federation run
Production Mode
Enable production mode (disables debug features)
Device
cpu
cuda
Compute device for training
Gpu Index
GPU index (e.g., '0', '1', or '0:3' for multi-GPU)
Random Seed
Random seed for reproducibility
Runtime Engine
torch
tensorflow
jax
Deep learning framework to use
Placement Group Strategy
SPREAD
PACK
STRICT_SPREAD
STRICT_PACK
Ray placement group strategy for resource allocation
🔗
Federated Learning Schema
Federated Learning Schema
TraditionalFederatedLearning
DecentralizedFederatedLearning
ClusterFederatedLearningSchema
Type of federated learning architecture
Federated Learning Topology
star
k_connect
ring
custom
Network topology for client connections
Draw Topology
Visualize the network topology
Adjacency Matrix File Name
Custom adjacency matrix file (for custom topology)
Client K Neighbors
Number of neighbors per client (for k_connect)
Client Role
train
test
eval
train-test
train-eval
test-eval
train-test-eval
Role of clients in the federation
🧠
Model Configuration
Model Type
cnn
resnet
vgg
transformer
mlp
lstm
gru
Type of neural network model
Transformer Model Size
small
base
medium
large
xlarge
Size of transformer model (if applicable)
Pretrained Models
Use pretrained model weights
📊
Dataset Configuration
Dataset Type
mnist
fmnist
cifar10
cifar100
emnist
svhn
imagenet
custom
Dataset for training
Data Distribution Kind
iid
non-iid
dirichlet
100
Data distribution across clients
Desired Distribution
Custom desired distribution (leave empty for null)
Dirichlet Beta
Dirichlet distribution parameter (lower = more heterogeneous)
Transform Input Size
Input image size after transformation
🎯
Training Parameters
Learning Rate
Learning rate for optimization
Loss Function
CrossEntropy
cross_entropy
nll_loss
mse_loss
bce_loss
Loss function for training
Optimizer
sgd
adam
adamw
rmsprop
Optimization algorithm
Weight Decay
Weight decay (L2 regularization), 0 for null/disabled
Number Of Epochs
Local epochs per federated round
Train Batch Size
Training batch size
Test Batch Size
Testing batch size
🌐
Federation Settings
Number Of Clients
Total number of federated clients
Client Sampling Rate
Fraction of clients participating per round
Federated Learning Rounds
Total federated learning rounds
Stop Avg Accuracy
Stop training when this accuracy is reached
🔄
Aggregation Strategy
Aggregation Strategy
FedAvg
fed_avg
fed_prox
fed_nova
scaffold
fed_adam
Aggregation algorithm
Fed Avg
true
false
simple
weighted
median
trimmed_mean
FedAvg variant or enable/disable
Distance Metric
cosine
euclidean
manhattan
Distance metric for model comparison
Distance Metric On Parameters
Compute distance on model parameters
Dynamic Sensitivity Percentage
Use dynamic sensitivity for aggregation
Sensitivity Percentage
Sensitivity percentage (send X% of most important chunks)
Remove Common Ids
Remove common IDs during aggregation
🎨
Clustering
Pre Computed Data Driven Clustering
Use pre-computed clustering
Do Cluster
Enable dynamic clustering
Clustering Period
Rounds between clustering updates
💾
Model Saving
Save Before Aggregation Models
Save models before aggregation
Save Global Models
Save global models after aggregation
Mean Accuracy To Csv
Export mean accuracy to CSV
🔒
Differential Privacy
Dp Enabled
Enable differential privacy
Dp Epsilon
Privacy budget (epsilon)
Dp Delta
Privacy parameter (delta)
Dp Clipping Norm
Gradient clipping norm
Dp Noise Multiplier
Noise multiplier for DP
🧩
Model Chunking
Chunking
Enable model chunking
Chunking With Gradients
Include gradients in chunks (MUST be true when chunking=true)
Chunking Parts
Number of chunks to divide model into
Chunking Random Section
Use random chunk selection (false = importance-based)
📈
Shapley Values
Shapley
Compute Shapley values for contribution
Shapley Type
value
cosine
coordinate
Shapley value computation method
📡
Ray Dashboard
Ray Dashboard
Enable Ray dashboard for monitoring
Ray Dashboard Port
Port for Ray dashboard
📊
Metrics Configuration
Metrics Round
Track round-level metrics
Metrics Memory
Track memory usage metrics
Metrics Performance
Track performance metrics
Metrics Communication
Track communication metrics
Metrics System
Track system metrics
Metrics Convergence
Track convergence metrics
Metrics Throughput
Track throughput metrics
Metrics Availability
Track availability metrics
📄
config.yaml Preview
YAML
●
Ready
Loading...
📋 Copy
⬇️ Download
💾 Save to Workspace
Saves config.yaml to your FedPilot directory
Configuration saved!