gpkg.mnist

gpkg.mnist contains logistic regression and CNN models for training with the MNIST dataset. These are used primarily for introducing Guild AI and for testing.

Name gpkg.mnist
Description CNN and multinomial logistic regression classifiers for MNIST digits (Guild AI)
Version 0.6.0.dev2
Source https://github.com/guildai/index/tree/master/gpkg/mnist
Author Guild AI

_check model

Operations

all

This operation does not have any flags.

cnn

This operation does not have any flags.

logreg

This operation does not have any flags.

cnn model

CNN classifier for MNIST.

Operations

evaluate

Evaluate a trained CNN.

This operation does not have any flags.

train

Train the CNN.
Flag Description Default
batch‑size Number of images to include in a training batch. 100
epochs Number of epochs to train. 5
learning‑rate Learning rate. 1e‑4

logreg model

Multinomial logistic regression classifier for MNIST.

Operations

evaluate

Evaluate a trained logistic regression.

This operation does not have any flags.

train

Train the logistic regression.
Flag Description Default
batch‑size Number of images to include in a training batch. 100
epochs Number of epochs to train. 5
learning‑rate Learning rate. 0.5

samples model

Sample MNIST images.

Operations

prepare

Generate a set of sample MNIST images.
Flag Description Default
count Number of images to generate. 100