ReferentialGym.modules package

Submodules

ReferentialGym.modules.batch_reshape_module module

ReferentialGym.modules.batch_reshape_module.build_BatchReshapeModule(id: str, config: Dict[str, object], input_stream_keys: List[str]) → ReferentialGym.modules.module.Module
class ReferentialGym.modules.batch_reshape_module.BatchReshapeModule(id: str, config: Dict[str, object], input_stream_keys: List[str])

Bases: ReferentialGym.modules.module.Module

compute(input_streams_dict: Dict[str, object]) → Dict[str, object]

Operates on inputs_dict that is made up of referents to the available stream. Make sure that accesses to its element are non-destructive.

Parameters

input_streams_dict – dict of str and data elements that follows self.input_stream_ids’s keywords and are extracted from self.input_stream_keys-named streams.

Returns

  • outputs_stream_dict:

ReferentialGym.modules.batch_reshape_repeat_module module

ReferentialGym.modules.batch_reshape_repeat_module.build_BatchReshapeRepeatModule(id: str, config: Dict[str, object], input_stream_keys: List[str]) → ReferentialGym.modules.module.Module
class ReferentialGym.modules.batch_reshape_repeat_module.BatchReshapeRepeatModule(id: str, config: Dict[str, object], input_stream_keys: List[str])

Bases: ReferentialGym.modules.module.Module

compute(input_streams_dict: Dict[str, object]) → Dict[str, object]

Operates on inputs_dict that is made up of referents to the available stream. Make sure that accesses to its element are non-destructive.

Parameters

input_streams_dict – dict of str and data elements that follows self.input_stream_ids’s keywords and are extracted from self.input_stream_keys-named streams.

Returns

  • outputs_stream_dict:

ReferentialGym.modules.concat_module module

ReferentialGym.modules.concat_module.build_ConcatModule(id: str, config: Dict[str, object], input_stream_keys: List[str]) → ReferentialGym.modules.module.Module
class ReferentialGym.modules.concat_module.ConcatModule(id: str, config: Dict[str, object], input_stream_keys: List[str])

Bases: ReferentialGym.modules.module.Module

compute(input_streams_dict: Dict[str, object]) → Dict[str, object]

Operates on inputs_dict that is made up of referents to the available stream. Make sure that accesses to its element are non-destructive.

Parameters

input_streams_dict – dict of str and data elements that follows self.input_stream_ids’s keywords and are extracted from self.input_stream_keys-named streams.

Returns

  • outputs_stream_dict:

ReferentialGym.modules.current_agent_module module

class ReferentialGym.modules.current_agent_module.CurrentAgentModule(id='current_agent', role=None)

Bases: ReferentialGym.modules.module.Module

set_ref(agent)
get_input_stream_ids()
clone(clone_id='a0')
save(path)
_tidyup()
_log(log_dict, batch_size)
register_hook(hook)
forward(sentences, experiences, multi_round=False, graphtype='straight_through_gumbel_softmax', tau0=0.2)
Parameters

  • sentences – Tensor of shape (batch_size, max_sentence_length, vocab_size) containing the padded sequence of (potentially one-hot-encoded) symbols.

  • experiences – Tensor of shape (batch_size, *self.obs_shape). Make sure to shuffle the experiences so that the order does not give away the target.

  • multi_round – Boolean defining whether to utter a sentence back or not.

  • graphtype – String defining the type of symbols used in the output sentence: - ‘categorical’: one-hot-encoded symbols. - ‘gumbel_softmax’: continuous relaxation of a categorical distribution. - ‘straight_through_gumbel_softmax’: improved continuous relaxation… - ‘obverter’: obverter training scheme…

  • tau0 – Float, temperature with which to apply gumbel-softmax estimator.

compute(input_streams_dict: Dict[str, object]) → Dict[str, object]

Compute the losses and return them along with the produced outputs.

Parameters

input_streams_dict –

Dict that should contain, at least, the following keys and values: - ‘sentences_logits’: Tensor of shape (batch_size, max_sentence_length, vocab_size) containing the padded sequence of logits over symbols. - ‘sentences_widx’: Tensor of shape (batch_size, max_sentence_length, 1) containing the padded sequence of symbols’ indices. - ‘sentences_one_hot’: Tensor of shape (batch_size, max_sentence_length, vocab_size) containing the padded sequence of one-hot-encoded symbols. - ‘experiences’: Tensor of shape (batch_size, *self.obs_shape). - ‘exp_latents’: Tensor of shape (batch_size, nbr_latent_dimensions). - ‘multi_round’: Boolean defining whether to utter a sentence back or not. - ‘graphtype’: String defining the type of symbols used in the output sentence:

  • ’categorical’: one-hot-encoded symbols.

  • ’gumbel_softmax’: continuous relaxation of a categorical distribution.

  • ’straight_through_gumbel_softmax’: improved continuous relaxation…

  • ’obverter’: obverter training scheme…

  • ’tau0’: Float, temperature with which to apply gumbel-softmax estimator.

  • ’sample’: Dict that contains the speaker and listener experiences as well as the target index.

  • ’config’: Dict of hyperparameters to the referential game.

  • ’mode’: String that defines what mode we are in, e.g. ‘train’ or ‘test’. Those keywords are expected.

  • ’it’: Integer specifying the iteration number of the current function call.

ReferentialGym.modules.dsprites_per_latent_accuracy_metric_module module

ReferentialGym.modules.dsprites_per_latent_accuracy_metric_module.build_dSpritesPerLatentAccuracyMetricModule(id: str, config: Dict[str, object], input_stream_ids: Dict[str, str] = None) → ReferentialGym.modules.module.Module
class ReferentialGym.modules.dsprites_per_latent_accuracy_metric_module.dSpritesPerLatentAccuracyMetricModule(id: str, config: Dict[str, object], input_stream_ids: Dict[str, str] = None)

Bases: ReferentialGym.modules.module.Module

compute(input_streams_dict: Dict[str, object]) → Dict[str, object]

ReferentialGym.modules.factor_vae_disentanglement_metric_module module

ReferentialGym.modules.factor_vae_disentanglement_metric_module.build_FactorVAEDisentanglementMetricModule(id: str, config: Dict[str, object], input_stream_ids: Dict[str, str] = None) → ReferentialGym.modules.module.Module
class ReferentialGym.modules.factor_vae_disentanglement_metric_module.FactorVAEDisentanglementMetricModule(id: str, config: Dict[str, object], input_stream_ids: Dict[str, str] = None)

Bases: ReferentialGym.modules.module.Module

_prune_dims(variances)

Mask for dimensions collapsed to the prior.

_generate_training_batch(dataset, model, batch_size, nbr_points, global_variances, active_dims)

Sample a set of training samples based on a batch of ground-truth data.

Args:

dataset: dataset to be sampled from. model: model that takes observations as input and

outputs a dim_representation sized representation for each observation.

batch_size: Number of points to be used to compute the training_sample. nbr_points: Number of points to be sampled for training/evaluation set. global_variances: Numpy vector with variances for all dimensions of

representation.

active_dims: Indexes of active dimensions.

Returns:

(num_factors, dim_representation)-sized numpy array with votes.

_generate_training_sample(dataset, model, batch_size, global_variances, active_dims)

Sample a single training sample based on a mini-batch of ground-truth data.

Args: dataset: dataset to be sampled from. model: model that takes observation as input and

outputs a representation.

batch_size: Number of points to be used to compute the training_sample. global_variances: Numpy vector with variances for all dimensions of

representation.

active_dims: Indexes of active dimensions.

Returns:

factor_index: Index of factor coordinate to be used. argmin: Index of representation coordinate with the least variance.

compute(input_streams_dict: Dict[str, object]) → Dict[str, object]

ReferentialGym.modules.flatten_module module

ReferentialGym.modules.flatten_module.build_FlattenModule(id: str, input_stream_keys: List[str]) → ReferentialGym.modules.module.Module
class ReferentialGym.modules.flatten_module.FlattenModule(id: str, input_stream_keys: List[str])

Bases: ReferentialGym.modules.module.Module

compute(input_streams_dict: Dict[str, object]) → Dict[str, object]

Operates on inputs_dict that is made up of referents to the available stream. Make sure that accesses to its element are non-destructive.

Parameters

input_streams_dict – dict of str and data elements that follows self.input_stream_ids’s keywords and are extracted from self.input_stream_keys-named streams.

Returns

  • outputs_stream_dict:

ReferentialGym.modules.grad_recorder_module module

ReferentialGym.modules.grad_recorder_module.build_GradRecorderModule(id: str, config: Dict[str, object] = None, input_stream_ids: Dict[str, str] = None) → ReferentialGym.modules.module.Module
class ReferentialGym.modules.grad_recorder_module.GradRecorderModule(id: str, config: Dict[str, object], input_stream_ids: Dict[str, str] = None)

Bases: ReferentialGym.modules.module.Module

compute(input_streams_dict: Dict[str, object]) → Dict[str, object]

ReferentialGym.modules.homoscedastic_multi_task_loss_module module

ReferentialGym.modules.homoscedastic_multi_task_loss_module.build_HomoscedasticMultiTasksLossModule(id: str, config: Dict[str, object], input_stream_ids: Dict[str, str] = None) → ReferentialGym.modules.module.Module
class ReferentialGym.modules.homoscedastic_multi_task_loss_module.HomoscedasticMultiTasksLossModule(id: str, config: Dict[str, object], input_stream_ids: Dict[str, str] = None)

Bases: ReferentialGym.modules.module.Module

compute(input_streams_dict: Dict[str, object]) → Dict[str, object]

Operates on inputs_dict that is made up of referents to the available stream. Make sure that accesses to its element are non-destructive.

Parameters

input_streams_dict – dict of str and data elements that follows self.input_stream_ids’s keywords and are extracted from self.input_stream_keys-named streams.

Returns

  • outputs_stream_dict:

ReferentialGym.modules.instantaneous_coordination_metric_module module

ReferentialGym.modules.instantaneous_coordination_metric_module.build_InstantaneousCoordinationMetricModule(id: str, config: Dict[str, object], input_stream_ids: Dict[str, str] = None) → ReferentialGym.modules.module.Module
class ReferentialGym.modules.instantaneous_coordination_metric_module.InstantaneousCoordinationMetricModule(id: str, config: Dict[str, object], input_stream_ids: Dict[str, str] = None)

Bases: ReferentialGym.modules.module.Module

compute(input_streams_dict: Dict[str, object]) → Dict[str, object]

ReferentialGym.modules.module module

class ReferentialGym.modules.module.Module(id: str, type: str, config: Dict[str, object], input_stream_ids: Dict[str, str])

Bases: torch.nn.modules.module.Module

get_id() → str
get_type() → str
get_input_stream_ids() → Dict[str, str]
compute(inputs_dict: Dict[str, object]) → Dict[str, object]

Operates on inputs_dict that is made up of referents to the available stream. Make sure that accesses to its element are non-destructive.

Parameters

inputs_dict – dict of str and data elements that follows self.input_stream_ids’s keywords and are extracted from self.input_stream_keys -named streams.

Returns

  • outputs_sream_dict:

ReferentialGym.modules.multi_head_classification_from_feature_map_module module

ReferentialGym.modules.multi_head_classification_from_feature_map_module.build_MultiHeadClassificationFromFeatureMapModule(id: str, config: Dict[str, object], input_stream_ids: Dict[str, str]) → ReferentialGym.modules.module.Module
class ReferentialGym.modules.multi_head_classification_from_feature_map_module.MultiHeadClassificationFromFeatureMapModule(id: str, config: Dict[str, object], heads: torch.nn.modules.container.ModuleList, input_stream_ids: Dict[str, str], final_fn: torch.nn.modules.module.Module = Softmax(dim=- 1))

Bases: ReferentialGym.modules.module.Module

compute(input_streams_dict: Dict[str, object]) → Dict[str, object]

Operates on inputs_dict that is made up of referents to the available stream. Make sure that accesses to its element are non-destructive.

Parameters

input_streams_dict – dict of str and data elements that follows self.input_stream_ids’s keywords and are extracted from self.input_stream_keys-named streams.

Returns

  • outputs_stream_dict:

ReferentialGym.modules.multi_head_classification_module module

ReferentialGym.modules.multi_head_classification_module.build_MultiHeadClassificationModule(id: str, config: Dict[str, object], input_stream_ids: Dict[str, str]) → ReferentialGym.modules.module.Module
class ReferentialGym.modules.multi_head_classification_module.MultiHeadClassificationModule(id: str, config: Dict[str, object], heads: torch.nn.modules.container.ModuleList, input_stream_ids: Dict[str, str], final_fn: torch.nn.modules.module.Module = Softmax(dim=- 1))

Bases: ReferentialGym.modules.module.Module

compute(input_streams_dict: Dict[str, object]) → Dict[str, object]

Operates on inputs_dict that is made up of referents to the available stream. Make sure that accesses to its element are non-destructive.

Parameters

input_streams_dict – dict of str and data elements that follows self.input_stream_ids’s keywords and are extracted from self.input_stream_keys-named streams.

Returns

  • outputs_stream_dict:

ReferentialGym.modules.multi_head_regression_module module

ReferentialGym.modules.multi_head_regression_module.build_MultiHeadRegressionModule(id: str, config: Dict[str, object], input_stream_ids: Dict[str, str]) → ReferentialGym.modules.module.Module
class ReferentialGym.modules.multi_head_regression_module.MultiHeadRegressionModule(id: str, config: Dict[str, object], heads: torch.nn.modules.container.ModuleList, input_stream_ids: Dict[str, str], final_fn: torch.nn.modules.module.Module = Softmax(dim=- 1))

Bases: ReferentialGym.modules.module.Module

compute(input_streams_dict: Dict[str, object]) → Dict[str, object]

Operates on inputs_dict that is made up of referents to the available stream. Make sure that accesses to its element are non-destructive.

Parameters

input_streams_dict – dict of str and data elements that follows self.input_stream_ids’s keywords and are extracted from self.input_stream_keys-named streams.

Returns

  • outputs_stream_dict:

ReferentialGym.modules.optimization_module module

ReferentialGym.modules.optimization_module.build_OptimizationModule(id: str, config: Dict[str, object], input_stream_ids: Dict[str, str] = None) → ReferentialGym.modules.module.Module
class ReferentialGym.modules.optimization_module.OptimizationModule(id: str, config: Dict[str, object], input_stream_ids: Dict[str, str] = None)

Bases: ReferentialGym.modules.module.Module

compute(input_streams_dict: Dict[str, object]) → Dict[str, object]

Operates on inputs_dict that is made up of referents to the available stream. Make sure that accesses to its element are non-destructive.

Parameters

input_streams_dict – dict of str and data elements that follows self.input_stream_ids’s keywords and are extracted from self.input_stream_keys-named streams.

Returns

  • outputs_stream_dict:

ReferentialGym.modules.per_epoch_logger_module module

ReferentialGym.modules.per_epoch_logger_module.build_PerEpochLoggerModule(id: str, config: Dict[str, object] = None, input_stream_ids: Dict[str, str] = None) → ReferentialGym.modules.module.Module
class ReferentialGym.modules.per_epoch_logger_module.PerEpochLoggerModule(id: str, config: Dict[str, object], input_stream_ids: Dict[str, str] = None)

Bases: ReferentialGym.modules.module.Module

compute(input_streams_dict: Dict[str, object]) → Dict[str, object]

ReferentialGym.modules.population_handler_module module

ReferentialGym.modules.population_handler_module.build_PopulationHandlerModule(id: str, prototype_speaker: ReferentialGym.modules.module.Module, prototype_listener: ReferentialGym.modules.module.Module, config: Dict[str, object], input_stream_ids: Dict[str, str]) → ReferentialGym.modules.module.Module
class ReferentialGym.modules.population_handler_module.PopulationHandlerModule(id: str, prototype_speaker: ReferentialGym.modules.module.Module, prototype_listener: ReferentialGym.modules.module.Module, config: Dict[str, object], input_stream_ids: Dict[str, str])

Bases: ReferentialGym.modules.module.Module

_select_agents()
bookkeeping(mode, epoch)
_reptile_step(learner, reptile_learner, nbr_grad_steps=1, verbose=False)
_apply_meta_update(meta_learner, meta_optimizer, learner)
compute(input_streams_dict: Dict[str, object]) → Dict[str, object]

Operates on inputs_dict that is made up of referents to the available stream. Make sure that accesses to its element are non-destructive.

Parameters

input_streams_dict – dict of str and data elements that follows self.input_stream_ids’s keywords and are extracted from self.input_stream_keys-named streams.

Returns

  • outputs_stream_dict:

ReferentialGym.modules.squeeze_module module

ReferentialGym.modules.squeeze_module.build_SqueezeModule(id: str, config: Dict[str, object], input_stream_keys: List[str]) → ReferentialGym.modules.module.Module
class ReferentialGym.modules.squeeze_module.SqueezeModule(id: str, config: Dict[str, object], input_stream_keys: List[str])

Bases: ReferentialGym.modules.module.Module

compute(input_streams_dict: Dict[str, object]) → Dict[str, object]

Operates on inputs_dict that is made up of referents to the available stream. Make sure that accesses to its element are non-destructive.

Parameters

input_streams_dict – dict of str and data elements that follows self.input_stream_ids’s keywords and are extracted from self.input_stream_keys-named streams.

Returns

  • outputs_stream_dict:

ReferentialGym.modules.topographic_similarity_metric_module module

ReferentialGym.modules.topographic_similarity_metric_module.build_TopographicSimilarityMetricModule(id: str, config: Dict[str, object], input_stream_ids: Dict[str, str] = None) → ReferentialGym.modules.module.Module
class ReferentialGym.modules.topographic_similarity_metric_module.TopographicSimilarityMetricModule(id: str, config: Dict[str, object], input_stream_ids: Dict[str, str] = None)

Bases: ReferentialGym.modules.module.Module

compute(input_streams_dict: Dict[str, object]) → Dict[str, object]

ReferentialGym.modules.visual_module module

ReferentialGym.modules.visual_module.build_VisualModule(id: str, config: Dict[str, object], input_stream_ids: Dict[str, str]) → ReferentialGym.modules.module.Module
class ReferentialGym.modules.visual_module.VisualModule(id, encoder, featout_converter, featout_normalization, config, input_stream_ids)

Bases: ReferentialGym.modules.module.Module

compute(input_streams_dict: Dict[str, object]) → Dict[str, object]

Operates on inputs_dict that is made up of referents to the available stream. Make sure that accesses to its element are non-destructive.

Parameters

input_streams_dict – dict of str and data elements that follows self.input_stream_ids’s keywords and are extracted from self.input_stream_keys-named streams.

Returns

  • outputs_stream_dict:

Module contents