export model

游雁

2023-02-13 b6b8b9cf8e3ade1202796c2a3a9e9e2e2aa9b325

export model

 funasr/runtime/python/onnxruntime/paraformer/rapid_paraformer/paraformer_onnx.py

@@ -9,11 +9,11 @@
import librosa
import numpy as np

from .utils.utils import (CharTokenizer, Hypothesis, ONNXRuntimeError,
from utils.utils import (CharTokenizer, Hypothesis, ONNXRuntimeError,
                    OrtInferSession, TokenIDConverter, get_logger,
                    read_yaml)
from .utils.postprocess_utils import sentence_postprocess
from .utils.frontend import WavFrontend
from utils.postprocess_utils import sentence_postprocess
from utils.frontend import WavFrontend

logging = get_logger()


 funasr/runtime/python/onnxruntime/paraformer/rapid_paraformer/utils/frontend.py

@@ -43,8 +43,6 @@
        opts.mel_opts.debug_mel = False
        self.opts = opts

        self.compute_fbank_feats = knf.OnlineFbank(self.opts)

        self.filter_length_min = filter_length_min
        self.filter_length_max = filter_length_max
        self.lfr_m = lfr_m
@@ -57,11 +55,12 @@
    def fbank(self,
              waveform: np.ndarray) -> Tuple[np.ndarray, np.ndarray]:
        waveform = waveform * (1 << 15)
        self.compute_fbank_feats.accept_waveform(self.opts.frame_opts.samp_freq, waveform.tolist())
        frames = self.compute_fbank_feats.num_frames_ready
        fbank_fn = knf.OnlineFbank(self.opts)
        fbank_fn.accept_waveform(self.opts.frame_opts.samp_freq, waveform.tolist())
        frames = fbank_fn.num_frames_ready
        mat = np.empty([frames, self.opts.mel_opts.num_bins])
        for i in range(frames):
            mat[i, :] = self.compute_fbank_feats.get_frame(i)
            mat[i, :] = fbank_fn.get_frame(i)
        feat = mat.astype(np.float32)
        feat_len = np.array(mat.shape[0]).astype(np.int32)
        return feat, feat_len

 kaldi-native-fbank

New file
@@ -0,0 +1 @@
../kaldi-native-fbank