Add Python API and Python examples for audio tagging (#753)

sherpa-onnx/python/csrc/CMakeLists.txt

@@ -1,6 +1,7 @@
 include_directories(${CMAKE_SOURCE_DIR})
 
 set(srcs
+  audio-tagging.cc
   circular-buffer.cc
   display.cc
   endpoint.cc

sherpa-onnx/python/csrc/audio-tagging.cc

@@ -0,0 +1,87 @@
+// sherpa-onnx/python/csrc/audio-tagging.cc
+//
+// Copyright (c)  2024  Xiaomi Corporation
+
+#include "sherpa-onnx/python/csrc/audio-tagging.h"
+
+#include <string>
+
+#include "sherpa-onnx/csrc/audio-tagging.h"
+
+namespace sherpa_onnx {
+
+static void PybindOfflineZipformerAudioTaggingModelConfig(py::module *m) {
+  using PyClass = OfflineZipformerAudioTaggingModelConfig;
+  py::class_<PyClass>(*m, "OfflineZipformerAudioTaggingModelConfig")
+      .def(py::init<>())
+      .def(py::init<const std::string &>(), py::arg("model"))
+      .def_readwrite("model", &PyClass::model)
+      .def("validate", &PyClass::Validate)
+      .def("__str__", &PyClass::ToString);
+}
+
+static void PybindAudioTaggingModelConfig(py::module *m) {
+  PybindOfflineZipformerAudioTaggingModelConfig(m);
+
+  using PyClass = AudioTaggingModelConfig;
+
+  py::class_<PyClass>(*m, "AudioTaggingModelConfig")
+      .def(py::init<>())
+      .def(py::init<const OfflineZipformerAudioTaggingModelConfig &, int32_t,
+                    bool, const std::string &>(),
+           py::arg("zipformer"), py::arg("num_threads") = 1,
+           py::arg("debug") = false, py::arg("provider") = "cpu")
+      .def_readwrite("zipformer", &PyClass::zipformer)
+      .def_readwrite("num_threads", &PyClass::num_threads)
+      .def_readwrite("debug", &PyClass::debug)
+      .def_readwrite("provider", &PyClass::provider)
+      .def("validate", &PyClass::Validate)
+      .def("__str__", &PyClass::ToString);
+}
+
+static void PybindAudioTaggingConfig(py::module *m) {
+  PybindAudioTaggingModelConfig(m);
+
+  using PyClass = AudioTaggingConfig;
+
+  py::class_<PyClass>(*m, "AudioTaggingConfig")
+      .def(py::init<>())
+      .def(py::init<const AudioTaggingModelConfig &, const std::string &,
+                    int32_t>(),
+           py::arg("model"), py::arg("labels"), py::arg("top_k") = 5)
+      .def_readwrite("model", &PyClass::model)
+      .def_readwrite("labels", &PyClass::labels)
+      .def_readwrite("top_k", &PyClass::top_k)
+      .def("validate", &PyClass::Validate)
+      .def("__str__", &PyClass::ToString);
+}
+
+static void PybindAudioEvent(py::module *m) {
+  using PyClass = AudioEvent;
+
+  py::class_<PyClass>(*m, "AudioEvent")
+      .def_property_readonly(
+          "name", [](const PyClass &self) -> std::string { return self.name; })
+      .def_property_readonly(
+          "index", [](const PyClass &self) -> int32_t { return self.index; })
+      .def_property_readonly(
+          "prob", [](const PyClass &self) -> float { return self.prob; })
+      .def("__str__", &PyClass::ToString);
+}
+
+void PybindAudioTagging(py::module *m) {
+  PybindAudioTaggingConfig(m);
+  PybindAudioEvent(m);
+
+  using PyClass = AudioTagging;
+
+  py::class_<PyClass>(*m, "AudioTagging")
+      .def(py::init<const AudioTaggingConfig &>(), py::arg("config"),
+           py::call_guard<py::gil_scoped_release>())
+      .def("create_stream", &PyClass::CreateStream,
+           py::call_guard<py::gil_scoped_release>())
+      .def("compute", &PyClass::Compute, py::arg("s"), py::arg("top_k") = -1,
+           py::call_guard<py::gil_scoped_release>());
+}
+
+}  // namespace sherpa_onnx

sherpa-onnx/python/csrc/audio-tagging.h

@@ -0,0 +1,16 @@
+// sherpa-onnx/python/csrc/audio-tagging.h
+//
+// Copyright (c)  2024  Xiaomi Corporation
+
+#ifndef SHERPA_ONNX_PYTHON_CSRC_AUDIO_TAGGING_H_
+#define SHERPA_ONNX_PYTHON_CSRC_AUDIO_TAGGING_H_
+
+#include "sherpa-onnx/python/csrc/sherpa-onnx.h"
+
+namespace sherpa_onnx {
+
+void PybindAudioTagging(py::module *m);
+
+}
+
+#endif  // SHERPA_ONNX_PYTHON_CSRC_AUDIO_TAGGING_H_

sherpa-onnx/python/csrc/offline-tts-vits-model-config.cc

@@ -16,7 +16,7 @@ void PybindOfflineTtsVitsModelConfig(py::module *m) {
   py::class_<PyClass>(*m, "OfflineTtsVitsModelConfig")
       .def(py::init<>())
       .def(py::init<const std::string &, const std::string &,
-                    const std::string &, const std::string, float, float,
+                    const std::string &, const std::string &, float, float,
                     float>(),
            py::arg("model"), py::arg("lexicon"), py::arg("tokens"),
            py::arg("data_dir") = "", py::arg("noise_scale") = 0.667,

sherpa-onnx/python/csrc/sherpa-onnx.cc

@@ -5,6 +5,7 @@
 #include "sherpa-onnx/python/csrc/sherpa-onnx.h"
 
 #include "sherpa-onnx/python/csrc/alsa.h"
+#include "sherpa-onnx/python/csrc/audio-tagging.h"
 #include "sherpa-onnx/python/csrc/circular-buffer.h"
 #include "sherpa-onnx/python/csrc/display.h"
 #include "sherpa-onnx/python/csrc/endpoint.h"
@@ -38,6 +39,7 @@ PYBIND11_MODULE(_sherpa_onnx, m) {
   m.doc() = "pybind11 binding of sherpa-onnx";
 
   PybindWaveWriter(&m);
+  PybindAudioTagging(&m);
 
   PybindFeatures(&m);
   PybindOnlineCtcFstDecoderConfig(&m);

sherpa-onnx/python/csrc/speaker-embedding-extractor.cc

@@ -14,7 +14,7 @@ static void PybindSpeakerEmbeddingExtractorConfig(py::module *m) {
   using PyClass = SpeakerEmbeddingExtractorConfig;
   py::class_<PyClass>(*m, "SpeakerEmbeddingExtractorConfig")
       .def(py::init<>())
-      .def(py::init<const std::string &, int32_t, bool, const std::string>(),
+      .def(py::init<const std::string &, int32_t, bool, const std::string &>(),
            py::arg("model"), py::arg("num_threads") = 1,
            py::arg("debug") = false, py::arg("provider") = "cpu")
       .def_readwrite("model", &PyClass::model)

sherpa-onnx/python/csrc/spoken-language-identification.cc

@@ -33,7 +33,7 @@ static void PybindSpokenLanguageIdentificationConfig(py::module *m) {
   py::class_<PyClass>(*m, "SpokenLanguageIdentificationConfig")
       .def(py::init<>())
       .def(py::init<const SpokenLanguageIdentificationWhisperConfig &, int32_t,
-                    bool, const std::string>(),
+                    bool, const std::string &>(),
            py::arg("whisper"), py::arg("num_threads") = 1,
            py::arg("debug") = false, py::arg("provider") = "cpu")
       .def_readwrite("whisper", &PyClass::whisper)
@@ -53,7 +53,7 @@ void PybindSpokenLanguageIdentification(py::module *m) {
            py::arg("config"), py::call_guard<py::gil_scoped_release>())
       .def("create_stream", &PyClass::CreateStream,
            py::call_guard<py::gil_scoped_release>())
-      .def("compute", &PyClass::Compute,
+      .def("compute", &PyClass::Compute, py::arg("s"),
            py::call_guard<py::gil_scoped_release>());
 }
 

sherpa-onnx/python/sherpa_onnx/__init__.py

@@ -1,5 +1,9 @@
 from _sherpa_onnx import (
     Alsa,
+    AudioEvent,
+    AudioTagging,
+    AudioTaggingConfig,
+    AudioTaggingModelConfig,
     CircularBuffer,
     Display,
     OfflineStream,
@@ -7,6 +11,7 @@ from _sherpa_onnx import (
     OfflineTtsConfig,
     OfflineTtsModelConfig,
     OfflineTtsVitsModelConfig,
+    OfflineZipformerAudioTaggingModelConfig,
     OnlineStream,
     SileroVadModelConfig,
     SpeakerEmbeddingExtractor,