Gender-Classifier-Mini · PNN 50 % · 1-bit

A Prime Neural Network (PNN) version of prithivMLmods/Gender-Classifier-Mini, made on primeneural.net. 74 linear layers are PNN layers at 50 % density: in each layer every output reads a fixed, structured set of inputs, the same number for every output, and no index is stored. Each connection is one bit, its sign, with an FP16 scale per input and per output.


Original model	prithivMLmods/Gender-Classifier-Mini
PNN layers	74 at 50 % density
Agreement with the original model (top-1, held-out data)	89.86 %
Distillation	30 epochs on HuggingFaceM4/FairFace (least-squares initialisation, hidden-state matching)
Model file	12.5 MB (1-bit)

Loading requires the PNN package (evaluation licence: info@primeneural.net). © PreciseStatistics GmbH (Prime Neural Networks) · Patent pending (US 19/685,168).

Downloads last month: -

Safetensors

Model size

9.01M params

Tensor type

F16

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for PrimeNeural/Gender-Classifier-Mini-pnn-d50-1bit-j30

Base model

google/siglip2-base-patch16-224

Finetuned

prithivMLmods/Gender-Classifier-Mini

Quantized

(2)

this model