| | |
|---|
| | | model_eb_file = os.path.join(model_dir, "model_eb.torchscripts") |
|---|
| | | |
|---|
| | | if not (os.path.exists(model_eb_file) and os.path.exists(model_bb_file)): |
|---|
| | | print(".onnx is not exist, begin to export onnx") |
|---|
| | | print(".onnx does not exist, begin to export onnx") |
|---|
| | | try: |
|---|
| | | from funasr import AutoModel |
|---|
| | | except: |
|---|
| | |
|---|
| | | ) -> List: |
|---|
| | | # make hotword list |
|---|
| | | hotwords, hotwords_length = self.proc_hotword(hotwords) |
|---|
| | | # import pdb; pdb.set_trace() |
|---|
| | | [bias_embed] = self.eb_infer(hotwords, hotwords_length) |
|---|
| | | [bias_embed] = self.eb_infer(torch.Tensor(hotwords), torch.Tensor(hotwords_length)) |
|---|
| | | # index from bias_embed |
|---|
| | | bias_embed = bias_embed.transpose(1, 0, 2) |
|---|
| | | _ind = np.arange(0, len(hotwords)).tolist() |
|---|
| | |
|---|
| | | try: |
|---|
| | | with torch.no_grad(): |
|---|
| | | if int(self.device_id) == -1: |
|---|
| | | outputs = self.ort_infer(feats, feats_len) |
|---|
| | | outputs = self.bb_infer(feats, feats_len) |
|---|
| | | am_scores, valid_token_lens = outputs[0], outputs[1] |
|---|
| | | else: |
|---|
| | | outputs = self.ort_infer(feats.cuda(), feats_len.cuda()) |
|---|
| | | outputs = self.bb_infer_infer(feats.cuda(), feats_len.cuda()) |
|---|
| | | am_scores, valid_token_lens = outputs[0].cpu(), outputs[1].cpu() |
|---|
| | | except: |
|---|
| | | # logging.warning(traceback.format_exc()) |
|---|
| | |
|---|
| | | return hotwords, hotwords_length |
|---|
| | | |
|---|
| | | def bb_infer( |
|---|
| | | self, feats: np.ndarray, feats_len: np.ndarray, bias_embed |
|---|
| | | self, feats, feats_len, bias_embed |
|---|
| | | ) -> Tuple[np.ndarray, np.ndarray]: |
|---|
| | | outputs = self.ort_infer_bb([feats, feats_len, bias_embed]) |
|---|
| | | return outputs |
|---|
| | | |
|---|
| | | def eb_infer(self, hotwords, hotwords_length): |
|---|
| | | outputs = self.ort_infer_eb([hotwords.astype(np.int32), hotwords_length.astype(np.int32)]) |
|---|
| | | outputs = self.ort_infer_eb([hotwords, hotwords_length]) |
|---|
| | | return outputs |
|---|
| | | |
|---|
| | | def decode(self, am_scores: np.ndarray, token_nums: int) -> List[str]: |
|---|
