| | |
|---|
| | | import math |
|---|
| | | from typing import Optional |
|---|
| | | import time |
|---|
| | | from funasr.utils.register import register_class, registry_tables |
|---|
| | | from funasr.register import tables |
|---|
| | | from funasr.datasets.audio_datasets.load_audio_extract_fbank import load_audio,extract_fbank |
|---|
| | | from funasr.utils.datadir_writer import DatadirWriter |
|---|
| | | from torch.nn.utils.rnn import pad_sequence |
|---|
| | |
|---|
| | | return int(self.frame_size_ms) |
|---|
| | | |
|---|
| | | |
|---|
| | | @register_class("model_classes", "FsmnVAD") |
|---|
| | | @tables.register("model_classes", "FsmnVAD") |
|---|
| | | class FsmnVAD(nn.Module): |
|---|
| | | """ |
|---|
| | | Author: Speech Lab of DAMO Academy, Alibaba Group |
|---|
| | |
|---|
| | | self.vad_opts.speech_to_sil_time_thres, |
|---|
| | | self.vad_opts.frame_in_ms) |
|---|
| | | |
|---|
| | | encoder_class = registry_tables.encoder_classes.get(encoder.lower()) |
|---|
| | | encoder_class = tables.encoder_classes.get(encoder.lower()) |
|---|
| | | encoder = encoder_class(**encoder_conf) |
|---|
| | | self.encoder = encoder |
|---|
| | | # init variables |
|---|
