| | |
|---|
| | | |
|---|
| | | return_raw_text = kwargs.get("return_raw_text", False) |
|---|
| | | # step.3 compute punc model |
|---|
| | | raw_text = None |
|---|
| | | if self.punc_model is not None: |
|---|
| | | if not len(result["text"].strip()): |
|---|
| | | if return_raw_text: |
|---|
| | | result["raw_text"] = "" |
|---|
| | | result["raw_text"] = raw_text = "" |
|---|
| | | else: |
|---|
| | | deep_update(self.punc_kwargs, cfg) |
|---|
| | | punc_res = self.inference( |
|---|
| | |
|---|
| | | if return_raw_text: |
|---|
| | | result["raw_text"] = raw_text |
|---|
| | | result["text"] = punc_res[0]["text"] |
|---|
| | | else: |
|---|
| | | raw_text = None |
|---|
| | | |
|---|
| | | # speaker embedding cluster after resorted |
|---|
| | | if self.spk_model is not None and kwargs.get("return_spk_res", True): |
|---|
