paddlespeech.cli.executor module

class paddlespeech.cli.executor.BaseExecutor(task: str, **kwargs)

Bases: ABC

An abstract executor of paddlespeech tasks.

Methods

__call__(*arg, **kwargs)	Python API to call an executor.
disable_task_loggers()	Disable all loggers in current task.
execute(argv)	Command line entry.
get_input_source(input_)	Get task input source from command line input.
infer(*args, **kwargs)	Model inference and put results into self._outputs.
postprocess(*args, **kwargs)	Output postprocess and return results.
preprocess(input, *args, **kwargs)	Input preprocess and return paddle.Tensor stored in self._inputs.
process_task_results(input_, results[, ...])	Handling task results and redirect stdout if needed.
show_rtf(info)	Calculate rft of current task and show results.

disable_task_loggers()

Disable all loggers in current task.

abstract execute(argv: List[str]) → bool

Command line entry. This method can only be accessed by a command line such as paddlespeech asr.

Args:

argv (List[str]): Arguments from command line.

Returns:

int: Result of the command execution. True for a success and False for a failure.

get_input_source(input_: Optional[Union[str, PathLike]]) → Dict[str, Union[str, PathLike]]

Get task input source from command line input.

Args:

input_ (Union[str, os.PathLike, None]): Input from command line.

Returns:

Dict[str, Union[str, os.PathLike]]: A dict with ids and inputs.

abstract infer(*args, **kwargs)

Model inference and put results into self._outputs. This method get input tensors from self._inputs, and write output tensors into self._outputs.

abstract postprocess(*args, **kwargs) → Union[str, PathLike]

Output postprocess and return results. This method get model output from self._outputs and convert it into human-readable results.

Returns:

Union[str, os.PathLike]: Human-readable results such as texts and audio files.

abstract preprocess(input: Any, *args, **kwargs)

Input preprocess and return paddle.Tensor stored in self._inputs. Input content can be a text(tts), a file(asr, cls), a stream(not supported yet) or anything needed.

Args:

input (Any): Input text/file/stream or other content.

process_task_results(input_: Optional[Union[str, PathLike]], results: Dict[str, PathLike], job_dump_result: bool = False)

Handling task results and redirect stdout if needed.

Args:

input_ (Union[str, os.PathLike, None]): Input from command line. results (Dict[str, os.PathLike]): Task outputs. job_dump_result (bool, optional): if True, dumps job results into file. Defaults to False.

show_rtf(info: Dict[str, List[float]])

Calculate rft of current task and show results.