Keras with Matplotlib - Jupyter Notebook

The jupyter.ipynb example demonstrates Trains automatic logging of code running in a Jupyter Notebook that uses Keras and Matplotlib. It trains a simple deep neural network on the Keras built-in MNIST dataset. It builds a sequential model using a categorical crossentropy loss objective function, specifies accuracy as the metric, and uses two callbacks: a TensorBoard callback and a model checkpoint callback. When the script runs, it creates an experiment named notebook example which is associated with the examples project.

Scalars

The loss and accuracy metric scalar plots appear in RESULTS > SCALARS, along with the resource utilization plots, which are titled :monitor: machine.

image

Plots

The example calls Matplotlib methods to create several sample plots, and TensorBoard methods to plot histograms for layer density. They appear in RESULTS > PLOTS.

image

image

image

Debug samples

Calls to Matplotlib methods log debug sample images. They appear in RESULTS > DEBUG SAMPLES.

image

Hyperparameters

Trains automatically logs TensorFlow DEFINEs. A parameter dictionary is logged by connecting it to the Task using a call to the Task.connect method.

task_params = {'num_scatter_samples': 60, 'sin_max_value': 20, 'sin_steps': 30}
task_params = task.connect(task_params)

Later in the Juputer Notebook, more parameters are added to the dictionary.

task_params['batch_size'] = 128
task_params['nb_classes'] = 10
task_params['nb_epoch'] = 6
task_params['hidden_dim'] = 512

Parameter dictionaries appear in CONFIGURATIONS > HYPER PARAMETERS > General.

image

The TensorFlow DEFINES appear in the TF_DEFINE subsection.

image

Log

Text printed to the console for training appears in RESULTS > LOG.

image

Artifacts

Model artifacts associated with the experiment appear in the experiment info panel (in the EXPERIMENTS tab), and in the model info panel (in the MODELS tab).

The experiment info panel shows model tracking, including the model name and design in ARTIFACTS > Output Model.

image

The model info panel contains the model details, including the model URL, framework, and snapshot locations.

image