trieste.models.utils#
This module contains auxiliary objects and functions that are used by multiple model types.
Module Contents#
- write_summary_data_based_metrics(dataset: trieste.data.Dataset, model: trieste.models.interfaces.ProbabilisticModel, prefix: str = '') None[source]#
Logging utility for writing TensorBoard summary of various metrics for model diagnostics.
- Parameters:
dataset – The dataset to use for computing the metrics. All available data in the dataset will be used.
model – The model to produce metrics for.
prefix – The prefix to add to “accuracy” category of model summaries.
- write_summary_kernel_parameters(kernel: gpflow.kernels.Kernel, prefix: str = '') None[source]#
Logging utility for writing TensorBoard summary of kernel parameters. Provides useful diagnostics for models with a GPflow kernel. Only trainable parameters are logged.
- Parameters:
kernel – The kernel to use for computing the metrics.
prefix – The prefix to add to “kernel” category of model summaries.
- write_summary_likelihood_parameters(likelihood: gpflow.likelihoods.Likelihood, prefix: str = '') None[source]#
Logging utility for writing TensorBoard summary of likelihood parameters. Provides useful diagnostics for models with a GPflow likelihood. Only trainable parameters are logged.
- Parameters:
likelihood – The likelihood to use for computing the metrics.
prefix – The prefix to add to “likelihood” category of model summaries.
- get_module_with_variables(model: trieste.models.interfaces.ProbabilisticModel, *dependencies: Any) tensorflow.Module[source]#
Return a fresh module with a model’s variables attached, which can then be extended with methods and saved using tf.saved_model.
- Parameters:
model – Model to extract variables from.
dependencies – Dependent objects whose variables should also be included.
- optimize_model_and_save_result(model: trieste.models.interfaces.TrainableProbabilisticModel, dataset: trieste.data.Dataset) None[source]#
Optimize the model objective and save the (optimizer-specific) optimization result in the model object. To access it, use
get_last_optimization_result.- Parameters:
dataset – The data with which to train the model.