Skip to main content

Module base

Module base 

Source
Expand description

Base types for machine learning.

Re-exports§

pub use precision::StorePrecisionSettings;

Modules§

data
Generic data structures and functions for training or inference.
helpers
Module for generic training and inference helpers.
model
Model definition for the Kord model.
precision
Precision selection for ML pipelines.

Structs§

KordItem
A single kord sample.
TrainConfig
The training configuration used for all training, inference, and hyper parameter tuning.

Constants§

DETERMINISTIC_GUESS_SIZE
The deterministic guess vector mirrors the MIDI note signature.
FREQUENCY_SPACE_SIZE
The standard frequency space size to use across all ML operations.
INPUT_SPACE_SIZE
The dimensionality of the sample tensor produced by kord_item_to_sample_tensor.
MEL_SPACE_SIZE
The standard mel space size to use across all ML operations.
NOTE_SIGNATURE_SIZE
The number of MIDI-indexed note bins used throughout the pipeline.
NUM_CLASSES
Backward-compatible alias for target dimensionality.
PITCH_CLASS_COUNT
The number of pitch classes in a single octave (C through B).
TARGET_SPACE_SIZE
The dimensionality of the target tensor produced by kord_item_to_target_tensor.