1 Performance Benchmarking
Test CPU vs GPU performance across different network types and activation functions
WebGPU Acceleration
GPU vs CPU benchmarking
Type-Generic
Float32, Int32, Uint32 networks
Network Growth
Self-evolving architectures
ADHD Evaluation
Advanced performance metrics
Test CPU vs GPU performance across different network types and activation functions
Single Forward Pass Timing
Each test runs one forward pass for precise GPU vs CPU measurement (like the working index.html demo)
Run a benchmark to see CPU vs GPU performance comparisons
About the Benchmark:
Create a neural network and watch it evolve through distributed micro-network optimization
Define the initial network structure
View your network architecture and track performance metrics
Create a network to see its structure
Total Layers
Parameters
Growth Events
ADHD Score
1. Micro-Network Surgery: Extract trainable sub-networks from checkpoint layers
2. Distributed Training: Parallel optimization across multiple worker threads
3. ADHD Evaluation: Performance scoring using Accuracy Deviation Heatmap Distribution
4. Network Evolution: Reintegrate improved architectures back into the main network