Spaces:
Running
on
T4
Running
on
T4
nithinraok
commited on
Commit
•
6d6c247
1
Parent(s):
27fad84
Update nemo_align.py
Browse files- nemo_align.py +4 -1
nemo_align.py
CHANGED
@@ -5,6 +5,7 @@ import torch
|
|
5 |
from nemo.utils import logging
|
6 |
from pathlib import Path
|
7 |
from viterbi_decoding import viterbi_decoding
|
|
|
8 |
|
9 |
BLANK_TOKEN = "<b>"
|
10 |
|
@@ -437,7 +438,9 @@ def get_start_end_for_segments(word_timestamps):
|
|
437 |
|
438 |
def align_tdt_to_ctc_timestamps(tdt_txt, model, audio_filepath):
|
439 |
if isinstance(model, EncDecHybridRNNTCTCModel):
|
440 |
-
|
|
|
|
|
441 |
else:
|
442 |
raise ValueError("Currently supporting hybrid models")
|
443 |
|
|
|
5 |
from nemo.utils import logging
|
6 |
from pathlib import Path
|
7 |
from viterbi_decoding import viterbi_decoding
|
8 |
+
from nemo.collections.asr.parts.submodules.ctc_decoding import CTCDecodingConfig
|
9 |
|
10 |
BLANK_TOKEN = "<b>"
|
11 |
|
|
|
438 |
|
439 |
def align_tdt_to_ctc_timestamps(tdt_txt, model, audio_filepath):
|
440 |
if isinstance(model, EncDecHybridRNNTCTCModel):
|
441 |
+
ctc_cfg = CTCDecodingConfig()
|
442 |
+
ctc_cfg.decoding = "greedy_batch"
|
443 |
+
model.change_decoding_strategy(decoding_cfg=ctc_cfg, decoder_type="ctc")
|
444 |
else:
|
445 |
raise ValueError("Currently supporting hybrid models")
|
446 |
|