Templates

talos.templates exposes dataset acquisition helpers, preset parameter dictionaries, model callbacks and complete Scan pipelines. Import it through talos.templates. These existing Python templates support education, testing and development; framework-specific SFD templates separately support CLI experiments.

Dataset acquisition is explicitly invoked by the caller. Dataset helpers read remote Autonomio CSVs, except MNIST, which uses the TensorFlow dataset loader and its download/cache behavior. Install TensorFlow for the preset parameter dictionaries, model callbacks and pipelines; dataset CSV helpers use the Talos core and network access.

Each category of templates consists at least assets based on four popular machine learning datasets:

In addition, some categories (e.g. datasets) include additional templates. These are listed below and can be accessed through the corresponding namespace without previous knowledge.

Datasets

Dataset conveniences are explicitly requested outside the experiment core. iris() downloads the Autonomio CSV, shuffles its rows and returns features with one-hot labels. Use scikit-learn's separate load_iris fixture in Scan when a bundled, network-independent dataset is required. Datasets are accessed through talos.templates.datasets. For example:

import talos
from sklearn.model_selection import train_test_split
x, y = talos.templates.datasets.iris()
x_train, x_val, y_train, y_val = train_test_split(x, y, test_size=.2, random_state=17)
assert x.shape == (150, 4) and y.shape == (150, 3)

Available datasets

breast_cancer
cervical_cancer
icu_mortality
telco_churn
titanic
iris
mnist

Parameter dictionaries

Params consist of an indicative and somewhat meaningful parameter space boundaries that can be used as the parameter dictionary for Scan() experiments. Parameter dictionaries are accessed through talos.templates.params. For example:

p = talos.templates.params.iris()
p.update({'epochs': [2], 'first_neuron': [8], 'hidden_layers': [0],
          'batch_size': [16], 'losses': ['categorical_crossentropy']})
assert 'optimizer' in p

Available parameter dictionaries

breast_cancer
cervical_cancer
titanic
iris

Model callbacks

Models consist of Keras models that can be used as an input model for Scan() experiments. Models are accessed through talos.templates.models. For example:

values = {name: choices[0] for name, choices in p.items()}
history, model = talos.templates.models.iris(x_train, y_train, x_val, y_val, values)
assert len(history.history['loss']) == 2

Available model callbacks

breast_cancer
cervical_cancer
titanic
iris

Complete pipelines

Pipelines are self-contained Scan() experiments where you simply execute the command and an experiment is performed. Pipelines are accessed through talos.templates.pipelines. For example:

scan_object = talos.templates.pipelines.iris(round_limit=1)
assert len(scan_object.data) == 1

Available pipelines

breast_cancer
cervical_cancer
titanic
iris

Return contracts and execution boundaries

Namespace	Public calls and outputs
`datasets`	`iris()`, `breast_cancer()`, `cervical_cancer()`, `titanic()`, `icu_mortality(samples=None)`, and `telco_churn(quantile=.5)` return `(x, y)`; Telco returns a list of two targets.
`datasets.mnist()`	Returns `(x_train, y_train, x_val, y_val)` with scaled image inputs and one-hot labels; prints the chosen input shape.
`params`	`iris()`, `breast_cancer()`, `cervical_cancer()` and `titanic(debug=False)` return candidate dictionaries with lists/range tuples and TensorFlow optimizer classes.
`models`	Each named model accepts `(x_train, y_train, x_val, y_val, params)` with scalar per-trial values and returns `(history, fitted_model)`. Calling it performs training.
`pipelines`	Each named pipeline accepts `round_limit=2, random_method='uniform_mersenne'`; Titanic also accepts `debug=False`. Calling it acquires data and runs Scan immediately.

Pipelines use the experiment name 'test' and create a new run folder. Even a one-trial pipeline can train for the preset epoch count; reduce epochs in an explicit parameter dictionary when a shorter fit is required. The page's Iris callback example uses an explicit two-epoch dictionary, while the pipeline example uses its own presets.

CSV helper sampling and row shuffling are not seeded through these function signatures. Titanic retains missing values and reports that condition. Telco's positive quantile transforms its two result targets into binary labels; a nonpositive quantile retains continuous values. Helpers do not establish train-only preprocessing or a final held-out scientific evaluation protocol for the caller.

Remote file availability, schema changes, missing values, filesystem/download failures and framework errors propagate. Do not treat preset architecture/parameter choices as validated model recommendations. Native Torch models require a caller-owned callback or the Torch SFD.

Datasets​

Available datasets​

Parameter dictionaries​

Available parameter dictionaries​

Model callbacks​

Available model callbacks​

Complete pipelines​

Available pipelines​

Return contracts and execution boundaries​

Read next​