i6_core.recognition.advanced_tree_search

class i6_core.recognition.advanced_tree_search.AdvancedTreeSearchJob(*args, **kwargs)

Parameters:

• crp – Common Rasr parameters for recognition

• feature_flow – Flow network for recognition

• feature_scorer – Feature scorer used in recognition. For AdvancedTreeSearchLmImageAndGlobalCacheJob RASR requires setting the feature scorer while actually not using it.

• search_parameters – Parameters for search e.g. beam-pruning, uses defaults defined below if not set

• lm_lookahead – Whether to perform language model lookahead or not

• lookahead_options – Options for the lm lookahead

• create_lattice – Recognizer option to produde lattices. Stored as FST

• eval_single_best – Extract the best path from the lattice

• eval_best_in_lattice –

• use_gpu – Flag to enable GPU decoding

• rtf – Expected rtf value to predict time requirement for the job

• mem – Memory requirement for the job

• cpu – CPU requirement for the job

• lmgc_mem – Memory requirement for the AdvancedTreeSearchLmImageAndGlobalCacheJob

• lmgc_alias – Alias for the AdvancedTreeSearchLmImageAndGlobalCacheJob

• lmgc_scorer – Dummy scorer for the AdvancedTreeSearchLmImageAndGlobalCacheJob which is required but unused

• model_combination_config – Configuration for model combination

• model_combination_post_config – Post config for model combination

• extra_config – Additional Config for recognition

• extra_post_config – Post config of additional config for recognition

cleanup_before_run(cmd, retry, task_id, *args)

classmethod create_config(crp: CommonRasrParameters, feature_flow: FlowNetwork, feature_scorer: FeatureScorer, search_parameters: Optional[Dict[str, Any]], lm_lookahead: bool, lookahead_options: Optional[Dict[str, Any]], create_lattice: bool, eval_single_best: bool, eval_best_in_lattice: bool, mem: float, cpu: int, lmgc_mem: float, lmgc_alias: Optional[str], lmgc_scorer: Optional[FeatureScorer], model_combination_config: Optional[RasrConfig], model_combination_post_config: Optional[RasrConfig], extra_config: Optional[RasrConfig], extra_post_config: Optional[RasrConfig], **kwargs)

create_files()

classmethod hash(kwargs)

Parameters:

parsed_args (dict[str]) –

Returns:

hash for job given the arguments

Return type:

str

run(task_id)

tasks()

Returns:

yields Task’s

Return type:

list[sisyphus.task.Task]

classmethod update_search_parameters(search_parameters)

class i6_core.recognition.advanced_tree_search.AdvancedTreeSearchLmImageAndGlobalCacheJob(*args, **kwargs)

cleanup_before_run(cmd, retry, *args)

classmethod create_config(crp, feature_scorer, extra_config, extra_post_config, **kwargs)

create_files()

classmethod find_arpa_lms(lm_config, lm_post_config=None)

classmethod hash(kwargs)

Parameters:

parsed_args (dict[str]) –

Returns:

hash for job given the arguments

Return type:

str

run()

tasks()

Returns:

yields Task’s

Return type:

list[sisyphus.task.Task]

class i6_core.recognition.advanced_tree_search.AdvancedTreeSearchWithRescoringJob(*args, **kwargs)

Parameters:

• crp – Common Rasr parameters for recognition

• feature_flow – Flow network for recognition

• feature_scorer – Feature scorer used in recognition. For AdvancedTreeSearchLmImageAndGlobalCacheJob RASR requires setting the feature scorer while actually not using it.

• search_parameters – Parameters for search e.g. beam-pruning, uses defaults defined below if not set

• lm_lookahead – Whether to perform language model lookahead or not

• lookahead_options – Options for the lm lookahead

• create_lattice – Recognizer option to produde lattices. Stored as FST

• eval_single_best – Extract the best path from the lattice

• eval_best_in_lattice –

• use_gpu – Flag to enable GPU decoding

• rtf – Expected rtf value to predict time requirement for the job

• mem – Memory requirement for the job

• cpu – CPU requirement for the job

• lmgc_mem – Memory requirement for the AdvancedTreeSearchLmImageAndGlobalCacheJob

• lmgc_alias – Alias for the AdvancedTreeSearchLmImageAndGlobalCacheJob

• lmgc_scorer – Dummy scorer for the AdvancedTreeSearchLmImageAndGlobalCacheJob which is required but unused

• model_combination_config – Configuration for model combination

• model_combination_post_config – Post config for model combination

• extra_config – Additional Config for recognition

• extra_post_config – Post config of additional config for recognition

classmethod create_config(rescorer_type, rescoring_lm_config, max_hypotheses, pruning_threshold, history_limit, rescoring_lookahead_scale, **kwargs)

class i6_core.recognition.advanced_tree_search.BidirectionalAdvancedTreeSearchJob(*args, **kwargs)

cleanup_before_run(cmd, retry, task_id, *args)

classmethod create_config(crp, feature_flow, feature_scorer, recognizer_parameters, search_parameters, lm_lookahead, lookahead_options, create_lattice, lattice_filter_type, eval_single_best, eval_best_in_lattice, mem, model_combination_config, model_combination_post_config, extra_config, extra_post_config, **kwargs)

create_files()

classmethod hash(kwargs)

Parameters:

parsed_args (dict[str]) –

Returns:

hash for job given the arguments

Return type:

str

run(task_id)

tasks()

Returns:

yields Task’s

Return type:

list[sisyphus.task.Task]

classmethod update_search_parameters(search_parameters)

class i6_core.recognition.advanced_tree_search.BuildGlobalCacheJob(*args, **kwargs)

Standalone job to create the global-cache for advanced-tree-search

Parameters:

• crp (rasr.CommonRasrParameters) – common RASR params (required: lexicon, acoustic_model, language_model, recognizer)

• extra_config (rasr.Configuration) – overlay config that influences the Job’s hash

• extra_post_config (rasr.Configuration) – overlay config that does not influences the Job’s hash

cleanup_before_run(cmd, retry, *args)

classmethod create_config(crp, extra_config, extra_post_config, **kwargs)

create_files()

classmethod hash(kwargs)

Parameters:

parsed_args (dict[str]) –

Returns:

hash for job given the arguments

Return type:

str

run()

tasks()

Returns:

yields Task’s

Return type:

list[sisyphus.task.Task]