Skip to content

Latest commit

 

History

History
35 lines (19 loc) · 509 Bytes

README.md

File metadata and controls

35 lines (19 loc) · 509 Bytes

NNToolkitCore

Cross-platform mobile Neural Network library for training and inference on device. CPU only.

On Apple platforms it uses Accelerate framework as a backend; On other platforms it uses Eigen for matmul, kissfft for dft, and ARM NEON intrinsics for acceleration.

NN layers: Conv1d GRU RNN LSTM BatchNorm Activation TimeDistributedDense Dense

DSP tools: Spectrogram Window functions

Algorithms were tested using tf 2.3.0

Examples and tests will be exposed in different repository.