MLP Explorer
ML expressivity vs. TinyXPU hardware efficiency ·
Source code
Depth (hidden layers)
2
Width (neurons/layer)
4
8
12
16
24
32
Array size
8×8
16×16
32×32
Task
Sum of sines
Square wave
TinyXPU Options
Double-buffer weights
Output taps at 8/12/16
ML Expressivity
in-browser training
Training…
Parameters
—
log₁₀ Loss (lower better)
—
Steps
—
Network Architecture
Hardware Comparison
TinyXPU vs TPU-like
Throughput
(MACs/cycle)
TinyXPU
—
TPU-like
—
Inference Latency
(cycles)
TinyXPU
—
TPU-like
—
Weight Load
Compute
Layer
Shape
Params
Adjust the controls above to explore the tradeoff.
PE Array Activity Visualization