ml
Layer Size Calculator
Build a network layer by layer — parameter counts, cumulative memory (MB), estimated max batch size given VRAM, network diagram SVG.
Architecture
Input
H1
H2
Out
Settings
Network Architecture
■ input■ hidden■ output
Totals
Total parameters109.4 K
Model size (float32)0.4173 MB
Recommended batch (VRAM)5235
Per-Layer Breakdown
LayerUnitsParamsCum.MB
Hidden 1128100.5 K100.5 K0.383
Hidden 2648.3 K108.7 K0.415
Output10650109.4 K0.417
params = (fan_in × units) + units (bias)
memory = params × bytes_per_dtype
batch ≈ ⌊VRAM × 0.8 / (model_size × 3)⌋
memory = params × bytes_per_dtype
batch ≈ ⌊VRAM × 0.8 / (model_size × 3)⌋