autogl.module.hpo

class autogl.module.hpo.AnnealAdvisorHPO(*args, **kwargs)[source]

Simulate anneal algorithm in advisor package See https://github.com/tobegit3hub/advisor for the package See .advisorbase.AdvisorBaseHPOptimizer for more information

classmethod build_hpo_from_args(args)[source]

Build a new hpo instance.

class autogl.module.hpo.AutoNE(*args, **kwargs)[source]

AutoNE HPOptimizer The Implementation of “AutoNE: Hyperparameter Optimization for Massive Network Embedding”(KDD 2019). See https://github.com/tadpole/AutoNE for more information

max_evals

int – The max rounds of evaluating HPs

subgraphs

int – The number of subgraphs

sub_evals

int – The number of evaluation times on each subgraph

sample_batch_size, sample_walk_length

int – Using for sampling subgraph, see torch_geometric.data.GraphSAINRandomWalkSampler

classmethod build_hpo_from_args(args)[source]

Build a new hpo instance.

optimize(trainer, dataset, time_limit=None, memory_limit=None)[source]

Optimize the HP by the method within give model and HP space

See .base.BaseHPOptimizer.optimize

class autogl.module.hpo.BayesAdvisor(*args, **kwargs)[source]

Bayes algorithm in advisor package See https://github.com/tobegit3hub/advisor for the package See .advisorbase.AdvisorBaseHPOptimizer for more information

classmethod build_hpo_from_args(args)[source]

Build a new hpo instance.

class autogl.module.hpo.CmaesAdvisorChoco(*args, **kwargs)[source]

CMAES algorithm in advisor package See https://github.com/tobegit3hub/advisor for the package See .advisorbase.AdvisorBaseHPOptimizer for more information

classmethod build_hpo_from_args(args)[source]

Build a new hpo instance.

class autogl.module.hpo.GridAdvisor(*args, **kwargs)[source]

Grid search algorithm in advisor package See https://github.com/tobegit3hub/advisor for the package See .advisorbase.AdvisorBaseHPOptimizer for more information

classmethod build_hpo_from_args(args)[source]

Build a new hpo instance.

class autogl.module.hpo.MocmaesAdvisorChoco(args)[source]

MOCMAES algorithm in advisor package See https://github.com/tobegit3hub/advisor for the package See .advisorbase.AdvisorBaseHPOptimizer for more information

classmethod build_hpo_from_args(args)[source]

Build a new hpo instance.

class autogl.module.hpo.QuasiAdvisorChoco(*args, **kwargs)[source]

Quasi random search algorithm in advisor package See https://github.com/tobegit3hub/advisor for the package See .advisorbase.AdvisorBaseHPOptimizer for more information

classmethod build_hpo_from_args(args)[source]

Build a new hpo instance.

class autogl.module.hpo.RandAdvisor(*args, **kwargs)[source]

Random search algorithm in advisor package See https://github.com/tobegit3hub/advisor for the package See .advisorbase.AdvisorBaseHPOptimizer for more information

classmethod build_hpo_from_args(args)[source]

Build a new hpo instance.

class autogl.module.hpo.TpeAdvisorHPO(*args, **kwargs)[source]

TPE algorithm in advisor package See https://github.com/tobegit3hub/advisor for the package See .advisorbase.AdvisorBaseHPOptimizer for more information

classmethod build_hpo_from_args(args)[source]

Build a new hpo instance.

autogl.module.hpo.build_hpo_from_name(name: str) → autogl.module.hpo.base.BaseHPOptimizer[source]
Parameters:name (str) – the name of hpo module.
Returns:the HPO built using default parameters
Return type:BaseHPOptimizer
Raises:AssertionError – If an invalid name is passed in