update repo

嘉渊

2023-06-15 ca666ab7d827c5aadf6866858b1da0acc9761563

update repo

 funasr/bin/sv_infer.py

@@ -1,35 +1,24 @@
# -*- encoding: utf-8 -*-
#!/usr/bin/env python3
# -*- encoding: utf-8 -*-
# Copyright FunASR (https://github.com/alibaba-damo-academy/FunASR). All Rights Reserved.
#  MIT License  (https://opensource.org/licenses/MIT)

import argparse
import logging
import os
import sys
from pathlib import Path
from typing import Any
from typing import List
from typing import Optional
from typing import Sequence
from typing import Tuple
from typing import Union

import numpy as np
import torch
from kaldiio import WriteHelper
from typeguard import check_argument_types
from typeguard import check_return_type

from funasr.utils.cli_utils import get_commandline_args
from funasr.tasks.sv import SVTask
from funasr.build_utils.build_model_from_file import build_model_from_file
from funasr.torch_utils.device_funcs import to_device
from funasr.torch_utils.set_all_random_seed import set_all_random_seed
from funasr.utils import config_argparse
from funasr.utils.types import str2bool
from funasr.utils.types import str2triple_str
from funasr.utils.types import str_or_none
from funasr.utils.misc import statistic_model_parameters


class Speech2Xvector:
    """Speech2Xvector class
@@ -56,10 +45,12 @@
        assert check_argument_types()

        # TODO: 1. Build SV model
        sv_model, sv_train_args = SVTask.build_model_from_file(
        sv_model, sv_train_args = build_model_from_file(
            config_file=sv_train_config,
            model_file=sv_model_file,
            device=device
            cmvn_file=None,
            device=device,
            task_name="sv"
        )
        logging.info("sv_model: {}".format(sv_model))
        logging.info("model parameter number: {}".format(statistic_model_parameters(sv_model)))
@@ -157,7 +148,3 @@
            kwargs.update(**d.download_and_unpack(model_tag))

        return Speech2Xvector(**kwargs)





 funasr/bin/sv_inference_launch.py

@@ -1,5 +1,5 @@
# -*- encoding: utf-8 -*-
#!/usr/bin/env python3
# -*- encoding: utf-8 -*-
# Copyright FunASR (https://github.com/alibaba-damo-academy/FunASR). All Rights Reserved.
#  MIT License  (https://opensource.org/licenses/MIT)

@@ -7,20 +7,6 @@
import logging
import os
import sys
from typing import Union, Dict, Any

from funasr.utils import config_argparse
from funasr.utils.cli_utils import get_commandline_args
from funasr.utils.types import str2bool
from funasr.utils.types import str2triple_str
from funasr.utils.types import str_or_none
import argparse
import logging
import os
import sys
from pathlib import Path
from typing import Any
from typing import List
from typing import Optional
from typing import Sequence
from typing import Tuple
@@ -30,18 +16,16 @@
import torch
from kaldiio import WriteHelper
from typeguard import check_argument_types
from typeguard import check_return_type

from funasr.utils.cli_utils import get_commandline_args
from funasr.tasks.sv import SVTask
from funasr.torch_utils.device_funcs import to_device
from funasr.bin.sv_infer import Speech2Xvector
from funasr.build_utils.build_streaming_iterator import build_streaming_iterator
from funasr.torch_utils.set_all_random_seed import set_all_random_seed
from funasr.utils import config_argparse
from funasr.utils.cli_utils import get_commandline_args
from funasr.utils.types import str2bool
from funasr.utils.types import str2triple_str
from funasr.utils.types import str_or_none
from funasr.utils.misc import statistic_model_parameters
from funasr.bin.sv_infer import Speech2Xvector


def inference_sv(
    output_dir: Optional[str] = None,
@@ -114,16 +98,15 @@
            data_path_and_name_and_type = [raw_inputs, "speech", "waveform"]
        
        # 3. Build data-iterator
        loader = SVTask.build_streaming_iterator(
            data_path_and_name_and_type,
        loader = build_streaming_iterator(
            task_name="sv",
            preprocess_args=None,
            data_path_and_name_and_type=data_path_and_name_and_type,
            dtype=dtype,
            batch_size=batch_size,
            key_file=key_file,
            num_workers=num_workers,
            preprocess_fn=None,
            collate_fn=None,
            allow_variable_data_keys=allow_variable_data_keys,
            inference=True,
            use_collate_fn=False,
        )
        
        # 7 .Start for-loop
@@ -173,8 +156,6 @@
    return _forward




def inference_launch(mode, **kwargs):
    if mode == "sv":
        return inference_sv(**kwargs)
@@ -182,6 +163,7 @@
        logging.info("Unknown decoding mode: {}".format(mode))
        return None


def get_parser():
    parser = config_argparse.ArgumentParser(
        description="Speaker Verification",