PyTorch sequential classification model example

An example of a sequential network that classifies digit images used as an OpenML flow. We use sub networks here in order to show the that network hierarchies can be achieved with ease.

import torch.nn
import torch.optim

import openml
import openml.extensions.pytorch
import openml.extensions.pytorch.layers

import logging

Enable logging in order to observe the progress while running the example.

openml.config.logger.setLevel(logging.DEBUG)
openml.extensions.pytorch.config.logger.setLevel(logging.DEBUG)

Define a sequential network that does the initial image reshaping and normalization model.

processing_net = torch.nn.Sequential(
    openml.extensions.pytorch.layers.Functional(function=torch.Tensor.reshape,
                                                shape=(-1, 1, 28, 28)),
    torch.nn.BatchNorm2d(num_features=1)
)

Define a sequential network that does the extracts the features from the image.

features_net = torch.nn.Sequential(
    torch.nn.Conv2d(in_channels=1, out_channels=32, kernel_size=5),
    torch.nn.LeakyReLU(),
    torch.nn.MaxPool2d(kernel_size=2),
    torch.nn.Conv2d(in_channels=32, out_channels=64, kernel_size=5),
    torch.nn.LeakyReLU(),
    torch.nn.MaxPool2d(kernel_size=2),
)

Define a sequential network that flattens the features and compiles the results into probabilities for each digit.

results_net = torch.nn.Sequential(
    openml.extensions.pytorch.layers.Functional(function=torch.Tensor.reshape,
                                                shape=(-1, 4 * 4 * 64)),
    torch.nn.Linear(in_features=4 * 4 * 64, out_features=256),
    torch.nn.LeakyReLU(),
    torch.nn.Dropout(),
    torch.nn.Linear(in_features=256, out_features=10),
)

The main network, composed of the above specified networks.

model = torch.nn.Sequential(
    processing_net,
    features_net,
    results_net
)

Download the OpenML task for the mnist 784 dataset.

task = openml.tasks.get_task(3573)

Run the model on the task (requires an API key).

run = openml.runs.run_model_on_task(model, task, avoid_duplicate_runs=False)
# Publish the experiment on OpenML (optional, requires an API key).
run.publish()

print('URL for run: %s/run/%d' % (openml.config.server, run.run_id))