i6_core.deprecated.returnn_search

class i6_core.deprecated.returnn_search.ReturnnSearchJob(*args, **kwargs)

Given a model checkpoint, run search task with RETURNN

Parameters:

• search_data (dict[str]) – dataset used for search

• model_checkpoint (Checkpoint) – TF model checkpoint. see ReturnnTrainingJob.

• returnn_config (ReturnnConfig) – object representing RETURNN config

• output_mode (str) – “txt” or “py”

• log_verbosity (int) – RETURNN log verbosity

• device (str) – RETURNN device, cpu or gpu

• time_rqmt (float|int) – job time requirement in hours

• mem_rqmt (float|int) – job memory requirement in GB

• cpu_rqmt (float|int) – job cpu requirement in GB

• returnn_python_exe (tk.Path|str|None) – path to the RETURNN executable (python binary or launch script)

• returnn_root (tk.Path|str|None) – path to the RETURNN src folder

create_files()

classmethod create_returnn_config(search_data, model_checkpoint, returnn_config, output_mode, log_verbosity, device, **kwargs)

Creates search RETURNN config :param dict[str] search_data: dataset used for search :param Checkpoint model_checkpoint: TF model checkpoint. see ReturnnTrainingJob. :param ReturnnConfig returnn_config: object representing RETURNN config :param str output_mode: “txt” or “py” :param int log_verbosity: RETURNN log verbosity :param str device: RETURNN device, cpu or gpu :rtype: ReturnnConfig

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]