EngineX-Iluvatar/enginex_bi_series-sherpa-onnx
Archived
16
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

Expose dither in python API (#2127)

Browse Source
This commit is contained in:
Nickolay V. Shmyrev
2025-04-17 11:47:48 +03:00
committed by GitHub
parent f3d23aa170
commit 84ed5d4288
1 changed files with 7 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
sherpa-onnx/python/sherpa_onnx/offline_recognizer.py
View File

@@ -49,6 +49,7 @@ class OfflineRecognizer(object):
num_threads: int = 1, num_threads: int = 1,
sample_rate: int = 16000, sample_rate: int = 16000,
feature_dim: int = 80, feature_dim: int = 80,
dither: float = 0.0,
decoding_method: str = "greedy_search", decoding_method: str = "greedy_search",
max_active_paths: int = 4, max_active_paths: int = 4,
hotwords_file: str = "", hotwords_file: str = "",
@@ -89,6 +90,11 @@ class OfflineRecognizer(object):
Sample rate of the training data used to train the model. Sample rate of the training data used to train the model.
feature_dim: feature_dim:
Dimension of the feature used to train the model. Dimension of the feature used to train the model.
dither:
Dithering constant (0.0 means no dither).
By default the audio samples are in range [-1,+1],
so dithering constant 0.00003 is a good value,
equivalent to the default 1.0 from kaldi
decoding_method: decoding_method:
Valid values: greedy_search, modified_beam_search. Valid values: greedy_search, modified_beam_search.
max_active_paths: max_active_paths:
@@ -142,6 +148,7 @@ class OfflineRecognizer(object):
feat_config = FeatureExtractorConfig( feat_config = FeatureExtractorConfig(
sampling_rate=sample_rate, sampling_rate=sample_rate,
feature_dim=feature_dim, feature_dim=feature_dim,
dither=dither,
) )
if len(hotwords_file) > 0 and decoding_method != "modified_beam_search": if len(hotwords_file) > 0 and decoding_method != "modified_beam_search":