speech_recognition/FunASR
1
0
Fork 0
mirror of https://github.com/modelscope/FunASR synced 2025-09-15 14:48:36 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

update

Browse Source
This commit is contained in:
游雁 2024-01-05 16:29:40 +08:00
commit 622d799de8
1 changed files with 6 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
funasr/models/monotonic_aligner/model.py
View File

@ -4,11 +4,13 @@ import torch
from torch.cuda.amp import autocast
from typing import Union, Dict, List, Tuple, Optional
from funasr.register import tables
from funasr.utils import postprocess_utils
from funasr.utils.datadir_writer import DatadirWriter
from funasr.models.paraformer.cif_predictor import mae_loss
from funasr.models.transformer.utils.add_sos_eos import add_sos_eos
from funasr.models.transformer.utils.nets_utils import make_pad_mask, pad_list
from funasr.metrics.compute_acc import th_accuracy
from funasr.train_utils.device_funcs import force_gatherable
from funasr.models.transformer.utils.add_sos_eos import add_sos_eos
from funasr.models.transformer.utils.nets_utils import make_pad_mask
from funasr.utils.timestamp_tools import ts_prediction_lfr6_standard
from funasr.utils import postprocess_utils
from funasr.utils.datadir_writer import DatadirWriter
@ -25,7 +27,6 @@ class MonotonicAligner(torch.nn.Module):
Achieving timestamp prediction while recognizing with non-autoregressive end-to-end ASR model
https://arxiv.org/abs/2301.12343
"""
def __init__(
self,
input_size: int = 80,
@ -41,7 +42,6 @@ class MonotonicAligner(torch.nn.Module):
length_normalized_loss: bool = False,
**kwargs,
):
super().__init__()
if specaug is not None:
@ -155,7 +155,6 @@ class MonotonicAligner(torch.nn.Module):
frontend=None,
**kwargs,
):
meta_data = {}
# extract fbank feats
time1 = time.perf_counter()
@ -190,8 +189,7 @@ class MonotonicAligner(torch.nn.Module):
timestamp_str, timestamp = ts_prediction_lfr6_standard(us_alpha[:encoder_out_lens[i] * 3],
us_peak[:encoder_out_lens[i] * 3],
copy.copy(token))
text_postprocessed, time_stamp_postprocessed, word_lists = postprocess_utils.sentence_postprocess(
token, timestamp)
text_postprocessed, time_stamp_postprocessed, _ = postprocess_utils.sentence_postprocess(token, timestamp)
result_i = {"key": key[i], "text": text_postprocessed,
"timestamp": time_stamp_postprocessed,
}