Skip to content

Latest commit

 

History

4 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

mantissa-nn

License Python Engine

The shared neural-net base for the mantissa family.

mantissa-nn holds the primitives every other package in the family reuses, so that the specialized packages depend on this base rather than on each other:

  • the C-engine binding (mantissa_nn._engine.engine) — loads mantissa once and gates on the dense primitive every model needs;
  • the pure-numpy reference backend (mantissa_nn._numpy_backend) — the identical call signatures in numpy, the correctness oracle;
  • the fully-connected layers Dense / Flatten and the Layer contract (mantissa_nn.layers).

Datasets are deliberately not here — each package owns and downloads its own, so they stay independent.

Built on top of it:

package adds
mantissa-cnn convolution + pooling layers, a CNN model zoo
mantissa-mlp the tabular training loop + tabular datasets
mantissa-autoencoder encoder/decoder model + decoder layers
from mantissa_nn import layers
from mantissa_nn._engine import engine     # the shared mantissa.Mantissa

The default compute path is the mantissa C engine (>= 0.2.1); backend="numpy" in the model packages selects the pure-numpy reference implementation.

The mantissa family

Part of the mantissa family: a low-precision engine written in C, with small Python packages built on top. Each package sits under the one it depends on — ⭐ marks where you are, and every other name links to its repo.

  • mantissa — low-precision neural-network engine in C (the core)
    • mantissa-perceptron — perceptron & ADALINE, the linear classics
    • mantissa-nn — shared neural-net primitives (layers, engine binding) (you are here)

Install

pip install mantissa-nn

For the development layout, clone the family repos as siblings and pip install -e . — a sibling mantissa/ checkout is picked up automatically when mantissa-core is not installed.

About

Shared neural-network primitives for the mantissa family on the mantissa C engine — engine binding, NumPy reference backend, and Dense/Flatten layers. Base package for cnn/mlp/auto-encoder. By Tekin Ertekin.

Topics

Resources

Contributing

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages