Changes for version 0.2610090 - 2026-10-09
- First release.
Modules
a small neural network, defined, trained and saved in Perl
the engines Peta::NN can compute on
Peta::NN on PDL ndarrays
Peta::NN on plain Perl arrays
Peta::NN on the graphics card
models put together into a model
characters to token indices, string pairs to edit labels
records with named fields, to train models on and measure them by
micro models fused into one, as they are, the seams inside
load a trained model and get answers from it
train a model until it meets given fidelity thresholds, at the smallest size that can
relu, tanh and sigmoid
fully connected layer
learned vectors for token indices
train a micro model from string pairs, export it as a model file
SGD with momentum, and Adam
independent pieces of work on several cores
models in series, and routed by a classifier
seeded random numbers that are the same on every perl