update

speech_asr

2023-03-15 e9f6703350fc6616b06c0e60944f6359a329e214

update

 funasr/bin/eend_ola_inference.py

@@ -227,8 +227,6 @@
            output_writer = open("{}/result.txt".format(output_path), "w")
        result_list = []
        for keys, batch in loader:
            logger.info("keys: {}".format(keys))
            logger.info("batch: {}".format(batch))
            assert isinstance(batch, dict), type(batch)
            assert all(isinstance(s, str) for s in keys), keys
            _bs = len(next(iter(batch.values())))

 funasr/models/frontend/wav_frontend.py

@@ -11,6 +11,8 @@
import funasr.models.frontend.eend_ola_feature as eend_ola_feature
from funasr.models.frontend.abs_frontend import AbsFrontend

from modelscope.utils.logger import get_logger
logger = get_logger()

def load_cmvn(cmvn_file):
    with open(cmvn_file, 'r', encoding='utf-8') as f:
@@ -485,6 +487,9 @@
        batch_size = input.size(0)
        feats = []
        feats_lens = []
        logger.info("batch_size: {}".format(batch_size))
        logger.info("input: {}".format(input))
        logger.info("input_lengths: {}".format(input_lengths))
        for i in range(batch_size):
            waveform_length = input_lengths[i]
            waveform = input[i][:waveform_length]