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

Add C++ runtime for Tele-AI/TeleSpeech-ASR (#970)

Browse Source
This commit is contained in:
Fangjun Kuang
2024-06-05 00:26:40 +08:00
committed by GitHub
parent f8dbc10146
commit fd5a0d1e00
52 changed files with 1052 additions and 145 deletions
Download Patch File Download Diff File Expand all files Collapse all files

41
sherpa-onnx/python/csrc/offline-model-config.cc
View File

@@ -29,25 +29,27 @@ void PybindOfflineModelConfig(py::module *m) {
using PyClass = OfflineModelConfig;
py::class_<PyClass>(*m, "OfflineModelConfig")
.def(py::init<const OfflineTransducerModelConfig &,
const OfflineParaformerModelConfig &,
const OfflineNemoEncDecCtcModelConfig &,
const OfflineWhisperModelConfig &,
const OfflineTdnnModelConfig &,
const OfflineZipformerCtcModelConfig &,
const OfflineWenetCtcModelConfig &, const std::string &,
int32_t, bool, const std::string &, const std::string &,
const std::string &, const std::string &>(),
py::arg("transducer") = OfflineTransducerModelConfig(),
py::arg("paraformer") = OfflineParaformerModelConfig(),
py::arg("nemo_ctc") = OfflineNemoEncDecCtcModelConfig(),
py::arg("whisper") = OfflineWhisperModelConfig(),
py::arg("tdnn") = OfflineTdnnModelConfig(),
py::arg("zipformer_ctc") = OfflineZipformerCtcModelConfig(),
py::arg("wenet_ctc") = OfflineWenetCtcModelConfig(),
py::arg("tokens"), py::arg("num_threads"), py::arg("debug") = false,
py::arg("provider") = "cpu", py::arg("model_type") = "",
py::arg("modeling_unit") = "cjkchar", py::arg("bpe_vocab") = "")
.def(
py::init<
const OfflineTransducerModelConfig &,
const OfflineParaformerModelConfig &,
const OfflineNemoEncDecCtcModelConfig &,
const OfflineWhisperModelConfig &, const OfflineTdnnModelConfig &,
const OfflineZipformerCtcModelConfig &,
const OfflineWenetCtcModelConfig &, const std::string &,
const std::string &, int32_t, bool, const std::string &,
const std::string &, const std::string &, const std::string &>(),
py::arg("transducer") = OfflineTransducerModelConfig(),
py::arg("paraformer") = OfflineParaformerModelConfig(),
py::arg("nemo_ctc") = OfflineNemoEncDecCtcModelConfig(),
py::arg("whisper") = OfflineWhisperModelConfig(),
py::arg("tdnn") = OfflineTdnnModelConfig(),
py::arg("zipformer_ctc") = OfflineZipformerCtcModelConfig(),
py::arg("wenet_ctc") = OfflineWenetCtcModelConfig(),
py::arg("telespeech_ctc") = "", py::arg("tokens"),
py::arg("num_threads"), py::arg("debug") = false,
py::arg("provider") = "cpu", py::arg("model_type") = "",
py::arg("modeling_unit") = "cjkchar", py::arg("bpe_vocab") = "")
.def_readwrite("transducer", &PyClass::transducer)
.def_readwrite("paraformer", &PyClass::paraformer)
.def_readwrite("nemo_ctc", &PyClass::nemo_ctc)
@@ -55,6 +57,7 @@ void PybindOfflineModelConfig(py::module *m) {
.def_readwrite("tdnn", &PyClass::tdnn)
.def_readwrite("zipformer_ctc", &PyClass::zipformer_ctc)
.def_readwrite("wenet_ctc", &PyClass::wenet_ctc)
.def_readwrite("telespeech_ctc", &PyClass::telespeech_ctc)
.def_readwrite("tokens", &PyClass::tokens)
.def_readwrite("num_threads", &PyClass::num_threads)
.def_readwrite("debug", &PyClass::debug)

65
sherpa-onnx/python/sherpa_onnx/offline_recognizer.py
View File

@@ -211,6 +211,71 @@ class OfflineRecognizer(object):
self.config = recognizer_config
return self
@classmethod
def from_telespeech_ctc(
cls,
model: str,
tokens: str,
num_threads: int = 1,
sample_rate: int = 16000,
feature_dim: int = 40,
decoding_method: str = "greedy_search",
debug: bool = False,
provider: str = "cpu",
):
"""
Please refer to
`<https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models>`_
to download pre-trained models.
Args:
model:
Path to ``model.onnx``.
tokens:
Path to ``tokens.txt``. Each line in ``tokens.txt`` contains two
columns::
symbol integer_id
num_threads:
Number of threads for neural network computation.
sample_rate:
Sample rate of the training data used to train the model. It is
ignored and is hard-coded in C++ to 40.
feature_dim:
Dimension of the feature used to train the model. It is ignored
and is hard-coded in C++ to 40.
decoding_method:
Valid values are greedy_search.
debug:
True to show debug messages.
provider:
onnxruntime execution providers. Valid values are: cpu, cuda, coreml.
"""
self = cls.__new__(cls)
model_config = OfflineModelConfig(
telespeech_ctc=model,
tokens=tokens,
num_threads=num_threads,
debug=debug,
provider=provider,
model_type="nemo_ctc",
)
feat_config = FeatureExtractorConfig(
sampling_rate=sample_rate,
feature_dim=feature_dim,
)
recognizer_config = OfflineRecognizerConfig(
feat_config=feat_config,
model_config=model_config,
decoding_method=decoding_method,
)
self.recognizer = _Recognizer(recognizer_config)
self.config = recognizer_config
return self
@classmethod
def from_nemo_ctc(
cls,