Skip to content

timsherwood/HardwareMatrixDecomp

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1 Commit
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Hardware Matrix Decomposition

Decompose neural network weight matrices into sets of smaller matrices, each operating in a weight-stationary fashion suitable for hardware accelerator execution.

Overview

Neural network inference relies on large matrix multiplications. This project explores decomposing those weight matrices into many smaller tiles that can each be mapped to a weight-stationary compute unit — keeping weights fixed in local storage while streaming activations through.

Setup

uv sync
uv sync --extra dev  # include dev tools

Development

uv run pytest
uv run ruff check .
uv run mypy src/

About

Decompose NN weight matrices into smaller tiles for weight-stationary hardware execution

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages