Merge 5fca61b1d5690d5470a2d20200c0cb67fd17d304 into 40cc0a4a2cf4b39d4c5405a55c7be14863287527

.gitignore

@@ -6,3 +6,4 @@ __pycache__/
 test-1hour.wav
 path.sh
 torch_version.py
+*.DS_Store

kaldifeat/csrc/CMakeLists.txt

@@ -7,6 +7,7 @@ set(kaldifeat_srcs
   feature-plp.cc
   feature-spectrogram.cc
   feature-window.cc
+  pitch-functions.cc
   matrix-functions.cc
   mel-computations.cc
   online-feature.cc

kaldifeat/csrc/pitch-functions.cc

@@ -0,0 +1,14 @@
+// kaldifeat/csrc/feature-fbank.cc
+//
+// Copyright (c)  2021  Xiaomi Corporation (authors: Fangjun Kuang)
+
+// This file is copied/modified from kaldi/src/feat/pitch-functions.cc
+
+#include "kaldifeat/csrc/pitch-functions.h"
+
+#include <cmath>
+
+namespace kaldifeat {
+
+
+}  // namespace kaldifeat

kaldifeat/csrc/pitch-functions.h

@@ -20,7 +20,8 @@
 
 #include <string>
 
-#include "torch/script.h"
+#include "kaldifeat/csrc/feature-common.h"
+#include "kaldifeat/csrc/feature-window.h"
 
 namespace kaldifeat {
 
@@ -144,10 +145,49 @@ struct PitchExtractionOptions {
     os << "nccf_ballast_online: " << nccf_ballast_online << "\n";
     os << "snip_edges: " << snip_edges << "\n";
     os << "device: " << device << "\n";
+    return os.str();
   }
 };
 
-// TODO(fangjun): Implement it
+
+std::ostream &operator<<(std::ostream &os, const PitchExtractionOptions &opts);
+
+class PitchComputer {
+ public:
+  using Options = PitchExtractionOptions;
+
+  explicit PitchComputer(const PitchExtractionOptions &opts);
+  ~PitchComputer();
+
+  PitchComputer &operator=(const PitchComputer &) = delete;
+  PitchComputer(const PitchComputer &) = delete;
+
+  int32_t Dim() const {
+    return 2;
+  }
+
+  // // if true, compute log_energy_pre_window but after dithering and dc removal
+  // bool NeedRawLogEnergy() const { return opts_.use_energy && opts_.raw_energy; }
+
+  // if true, compute log_energy_pre_window but after dithering and dc removal
+  bool NeedRawLogEnergy() const;
+
+  // const FrameExtractionOptions &GetFrameOptions() const {
+  //   return opts_.frame_opts;
+  // }
+
+  const FrameExtractionOptions &GetFrameOptions();
+
+  const PitchExtractionOptions &GetOptions() const { return opts_; }
+
+  torch::Tensor Compute(torch::Tensor signal_raw_log_energy, float vtln_warp,
+                        const torch::Tensor &signal_frame);
+
+ private:
+  PitchExtractionOptions opts_;
+};
+
+using Pitch = OfflineFeatureTpl<PitchComputer>;
 
 }  // namespace kaldifeat
 

kaldifeat/python/csrc/CMakeLists.txt

@@ -5,6 +5,7 @@ pybind11_add_module(_kaldifeat
   feature-plp.cc
   feature-spectrogram.cc
   feature-window.cc
+  pitch-functions.cc
   kaldifeat.cc
   mel-computations.cc
   online-feature.cc

kaldifeat/python/csrc/pitch-functions.cc

@@ -0,0 +1,133 @@
+// kaldifeat/python/csrc/pitch-functions.cc
+//
+// Copyright (c)  2024             (authors: Feiteng Li)
+
+#include "kaldifeat/python/csrc/pitch-functions.h"
+
+#include <memory>
+#include <string>
+
+#include "kaldifeat/csrc/pitch-functions.h"
+#include "kaldifeat/python/csrc/utils.h"
+
+namespace kaldifeat {
+
+static void PybindPitchOptions(py::module &m) {
+  using PyClass = PitchExtractionOptions;
+  py::class_<PyClass>(m, "PitchOptions")
+      .def(py::init<>())
+      .def(py::init([](float samp_freq = 16000,
+                       float frame_shift_ms = 10.0, float frame_length_ms = 25.0, float preemph_coeff = 0.0,
+                       float min_f0 = 50, float max_f0 = 400, float soft_min_f0 = 10.0,
+                       float penalty_factor = 0.1, float lowpass_cutoff = 1000, float resample_freq = 4000,
+                       float delta_pitch = 0.005, float nccf_ballast = 7000,
+                       int32_t lowpass_filter_width = 1, int32_t upsample_filter_width = 5,
+                       int32_t max_frames_latency = 0, int32_t frames_per_chunk = 0,
+                       bool simulate_first_pass_online = false, int32_t recompute_frame = 500,
+                       bool nccf_ballast_online = false, bool snip_edges = true,
+                       py::object device =
+                           py::str("cpu")) -> std::unique_ptr<PitchExtractionOptions> {
+             auto opts = std::make_unique<PitchExtractionOptions>();
+             opts->samp_freq = samp_freq;
+              opts->frame_shift_ms = frame_shift_ms;
+              opts->frame_length_ms = frame_length_ms;
+              opts->preemph_coeff = preemph_coeff;
+              opts->min_f0 = min_f0;
+              opts->max_f0 = max_f0;
+              opts->soft_min_f0 = soft_min_f0;
+              opts->penalty_factor = penalty_factor;
+              opts->lowpass_cutoff = lowpass_cutoff;
+              opts->resample_freq = resample_freq;
+              opts->delta_pitch = delta_pitch;
+              opts->nccf_ballast = nccf_ballast;
+              opts->lowpass_filter_width = lowpass_filter_width;
+              opts->upsample_filter_width = upsample_filter_width;
+              opts->max_frames_latency = max_frames_latency;
+              opts->frames_per_chunk = frames_per_chunk;
+              opts->simulate_first_pass_online = simulate_first_pass_online;
+              opts->recompute_frame = recompute_frame;
+              opts->nccf_ballast_online = nccf_ballast_online;
+              opts->snip_edges = snip_edges;
+
+             std::string s = static_cast<py::str>(device);
+             opts->device = torch::Device(s);
+
+             return opts;
+           }),
+          py::arg("samp_freq") = 16000, py::arg("frame_shift_ms") = 10.0,
+          py::arg("frame_length_ms") = 25.0, py::arg("preemph_coeff") = 0.0,
+          py::arg("min_f0") = 50, py::arg("max_f0") = 400, py::arg("soft_min_f0") = 10.0,
+          py::arg("penalty_factor") = 0.1, py::arg("lowpass_cutoff") = 1000, py::arg("resample_freq") = 4000,
+          py::arg("delta_pitch") = 0.005, py::arg("nccf_ballast") = 7000,
+          py::arg("lowpass_filter_width") = 1, py::arg("upsample_filter_width") = 5,
+          py::arg("max_frames_latency") = 0, py::arg("frames_per_chunk") = 0,
+          py::arg("simulate_first_pass_online") = false, py::arg("recompute_frame") = 500,
+          py::arg("nccf_ballast_online") = false, py::arg("snip_edges") = true,
+          py::arg("device") = py::str("cpu"))
+
+      .def_readwrite("samp_freq", &PyClass::samp_freq)
+      .def_readwrite("frame_shift_ms", &PyClass::frame_shift_ms)
+      .def_readwrite("frame_length_ms", &PyClass::frame_length_ms)
+      .def_readwrite("preemph_coeff", &PyClass::preemph_coeff)
+      .def_readwrite("min_f0", &PyClass::min_f0)
+      .def_readwrite("max_f0", &PyClass::max_f0)
+      .def_readwrite("soft_min_f0", &PyClass::soft_min_f0)
+      .def_readwrite("penalty_factor", &PyClass::penalty_factor)
+      .def_readwrite("lowpass_cutoff", &PyClass::lowpass_cutoff)
+      .def_readwrite("resample_freq", &PyClass::resample_freq)
+      .def_readwrite("delta_pitch", &PyClass::delta_pitch)
+      .def_readwrite("nccf_ballast", &PyClass::nccf_ballast)
+      .def_readwrite("lowpass_filter_width", &PyClass::lowpass_filter_width)
+      .def_readwrite("upsample_filter_width", &PyClass::upsample_filter_width)
+      .def_readwrite("max_frames_latency", &PyClass::max_frames_latency)
+      .def_readwrite("frames_per_chunk", &PyClass::frames_per_chunk)
+      .def_readwrite("simulate_first_pass_online",
+                     &PyClass::simulate_first_pass_online)
+      .def_readwrite("recompute_frame", &PyClass::recompute_frame)
+      .def_readwrite("nccf_ballast_online", &PyClass::nccf_ballast_online)
+      .def_readwrite("snip_edges", &PyClass::snip_edges)
+      .def_property(
+          "device",
+          [](const PyClass &self) -> py::object {
+            py::object ans = py::module_::import("torch").attr("device");
+            return ans(self.device.str());
+          },
+          [](PyClass &self, py::object obj) -> void {
+            std::string s = static_cast<py::str>(obj);
+            self.device = torch::Device(s);
+          })
+      .def("__str__",
+           [](const PyClass &self) -> std::string { return self.ToString(); })
+      .def("as_dict",
+           [](const PyClass &self) -> py::dict { return AsDict(self); })
+      .def_static(
+          "from_dict",
+          [](py::dict dict) -> PyClass { return PitchOptionsFromDict(dict); })
+      .def(py::pickle(
+          [](const PyClass &self) -> py::dict { return AsDict(self); },
+          [](py::dict dict) -> PyClass { return PitchOptionsFromDict(dict); }));
+}
+
+static void PybindPitch(py::module &m) {
+  using PyClass = Pitch;
+  py::class_<PyClass>(m, "Pitch")
+      .def(py::init<const PitchOptions &>(), py::arg("opts"))
+      .def("dim", &PyClass::Dim)
+      .def_property_readonly("options", &PyClass::GetOptions)
+      .def("compute_features", &PyClass::ComputeFeatures, py::arg("wave"),
+           py::arg("vtln_warp"), py::call_guard<py::gil_scoped_release>())
+      .def(py::pickle(
+          [](const PyClass &self) -> py::dict {
+            return AsDict(self.GetOptions());
+          },
+          [](py::dict dict) -> std::unique_ptr<PyClass> {
+            return std::make_unique<PyClass>(PitchOptionsFromDict(dict));
+          }));
+}
+
+void PybindPitchFunctions(py::module &m) {
+  PybindPitchOptions(m);
+  PybindPitch(m);
+}
+
+}  // namespace kaldifeat

kaldifeat/python/csrc/pitch-functions.h

@@ -0,0 +1,16 @@
+// kaldifeat/python/csrc/pitch-functions.h
+//
+// Copyright (c)  2024             (authors: Feiteng Li)
+
+#ifndef KALDIFEAT_PYTHON_CSRC_PITCH_FUNCTIONS_H_
+#define KALDIFEAT_PYTHON_CSRC_PITCH_FUNCTIONS_H_
+
+#include "kaldifeat/python/csrc/kaldifeat.h"
+
+namespace kaldifeat {
+
+void PybindPitchFunctions(py::module &m);
+
+}  // namespace kaldifeat
+
+#endif  // KALDIFEAT_PYTHON_CSRC_PITCH_FUNCTIONS_H_

kaldifeat/python/csrc/utils.cc

@@ -152,6 +152,69 @@ py::dict AsDict(const WhisperFbankOptions &opts) {
   return dict;
 }
 
+
+PitchOptions PitchOptionsFromDict(py::dict dict) {
+  PitchOptions opts;
+
+  FROM_DICT(float_, samp_freq);
+  FROM_DICT(float_, frame_shift_ms);
+  FROM_DICT(float_, frame_length_ms);
+  FROM_DICT(float_, preemph_coeff);
+  FROM_DICT(float_, min_f0);
+  FROM_DICT(float_, max_f0);
+  FROM_DICT(float_, soft_min_f0);
+  FROM_DICT(float_, penalty_factor);
+  FROM_DICT(float_, lowpass_cutoff);
+  FROM_DICT(float_, resample_freq);
+  FROM_DICT(float_, delta_pitch);
+  FROM_DICT(float_, nccf_ballast);
+  FROM_DICT(int_, lowpass_filter_width);
+  FROM_DICT(int_, upsample_filter_width);
+  FROM_DICT(int_, max_frames_latency);
+  FROM_DICT(int_, frames_per_chunk);
+  FROM_DICT(bool_, simulate_first_pass_online);
+  FROM_DICT(int_, recompute_frame);
+  FROM_DICT(bool_, nccf_ballast_online);
+  FROM_DICT(bool_, snip_edges);
+
+  if (dict.contains("device")) {
+    opts.device = torch::Device(std::string(py::str(dict["device"])));
+  }
+
+  return opts;
+}
+
+py::dict AsDict(const PitchOptions &opts) {
+  py::dict dict;
+
+  AS_DICT(samp_freq);
+  AS_DICT(frame_shift_ms);
+  AS_DICT(frame_length_ms);
+  AS_DICT(preemph_coeff);
+  AS_DICT(min_f0);
+  AS_DICT(max_f0);
+  AS_DICT(soft_min_f0);
+  AS_DICT(penalty_factor);
+  AS_DICT(lowpass_cutoff);
+  AS_DICT(resample_freq);
+  AS_DICT(delta_pitch);
+  AS_DICT(nccf_ballast);
+  AS_DICT(lowpass_filter_width);
+  AS_DICT(upsample_filter_width);
+  AS_DICT(max_frames_latency);
+  AS_DICT(frames_per_chunk);
+  AS_DICT(simulate_first_pass_online);
+  AS_DICT(recompute_frame);
+  AS_DICT(nccf_ballast_online);
+  AS_DICT(snip_edges);
+
+  auto torch_device = py::module_::import("torch").attr("device");
+  dict["device"] = torch_device(opts.device.str());
+
+  return dict;
+}
+
+
 MfccOptions MfccOptionsFromDict(py::dict dict) {
   MfccOptions opts;
 

kaldifeat/python/csrc/utils.h

@@ -10,6 +10,7 @@
 #include "kaldifeat/csrc/feature-plp.h"
 #include "kaldifeat/csrc/feature-spectrogram.h"
 #include "kaldifeat/csrc/feature-window.h"
+#include "kaldifeat/csrc/pitch-functions.h"
 #include "kaldifeat/csrc/mel-computations.h"
 #include "kaldifeat/csrc/whisper-fbank.h"
 #include "kaldifeat/python/csrc/kaldifeat.h"
@@ -26,8 +27,12 @@
  * all fields use their default values.
  */
 
+
 namespace kaldifeat {
 
+using PitchOptions = kaldifeat::PitchExtractionOptions;
+
+
 FrameExtractionOptions FrameExtractionOptionsFromDict(py::dict dict);
 py::dict AsDict(const FrameExtractionOptions &opts);
 
@@ -40,6 +45,9 @@ py::dict AsDict(const FbankOptions &opts);
 WhisperFbankOptions WhisperFbankOptionsFromDict(py::dict dict);
 py::dict AsDict(const WhisperFbankOptions &opts);
 
+PitchOptions PitchOptionsFromDict(py::dict dict);
+py::dict AsDict(const PitchOptions &opts);
+
 MfccOptions MfccOptionsFromDict(py::dict dict);
 py::dict AsDict(const MfccOptions &opts);
 

kaldifeat/python/kaldifeat/__init__.py

@@ -15,6 +15,7 @@ from _kaldifeat import (
     FrameExtractionOptions,
     MelBanksOptions,
     MfccOptions,
+    PitchOptions,
     PlpOptions,
     SpectrogramOptions,
     WhisperFbankOptions,
@@ -25,9 +26,21 @@ from .fbank import Fbank, OnlineFbank
 from .mfcc import Mfcc, OnlineMfcc
 from .offline_feature import OfflineFeature
 from .online_feature import OnlineFeature
+from .pitch import Pitch
 from .plp import OnlinePlp, Plp
 from .spectrogram import Spectrogram
 from .whisper_fbank import WhisperFbank
 
 cmake_prefix_path = _Path(__file__).parent / "share" / "cmake"
 del _Path
+
+__version__ = '1.25.4.dev20240124+cpu.torch2.1.1'
+__version__ = '1.25.4.dev20240124+cpu.torch2.1.1'
+__version__ = '1.25.4.dev20240124+cpu.torch2.1.1'
+__version__ = '1.25.4.dev20240124+cpu.torch2.1.1'
+__version__ = '1.25.4.dev20240124+cpu.torch2.1.1'
+__version__ = '1.25.4.dev20240124+cpu.torch2.1.1'
+__version__ = '1.25.4.dev20240124+cpu.torch2.1.1'
+__version__ = '1.25.4.dev20240124+cpu.torch2.1.1'
+__version__ = '1.25.4.dev20240124+cpu.torch2.1.1'
+__version__ = '1.25.4.dev20240124+cpu.torch2.1.1'

kaldifeat/python/kaldifeat/pitch.py

@@ -0,0 +1,13 @@
+# Copyright (c)  2024             (authors: Feiteng Li)
+
+
+import _kaldifeat
+
+from .offline_feature import OfflineFeature
+
+
+class Pitch(OfflineFeature):
+    def __init__(self, opts: _kaldifeat.FbankOptions):
+        super().__init__(opts)
+        self.computer = _kaldifeat.Pitch(opts)
+

kaldifeat/python/tests/CMakeLists.txt

@@ -23,6 +23,7 @@ set(py_test_files
   test_mel_bank_options.py
   test_mfcc.py
   test_mfcc_options.py
+  test_pitch.py
   test_plp.py
   test_plp_options.py
   test_spectrogram.py

kaldifeat/python/tests/test_pitch.py

@@ -0,0 +1,41 @@
+#!/usr/bin/env python3
+
+# Copyright      2021-2022  Xiaomi Corporation (authors: Fangjun Kuang)
+
+import pickle
+from pathlib import Path
+
+import torch
+from utils import get_devices, read_ark_txt, read_wave
+
+import kaldifeat
+
+cur_dir = Path(__file__).resolve().parent
+
+
+def test_pitch_default():
+    print("=====test_pitch_default=====")
+    filename = cur_dir / "test_data/test.wav"
+    wave = read_wave(filename)
+    gt = read_ark_txt(cur_dir / "test_data/test-pitch.txt")
+
+    cpu_features = None
+    for device in get_devices():
+        print("device", device)
+        opts = kaldifeat.PitchOptions()
+        opts.device = device
+        opts.frame_opts.dither = 0
+        pitch = kaldifeat.Pitch(opts)
+
+        features = pitch(wave)
+        assert features.device.type == "cpu"
+        assert torch.allclose(features, gt, rtol=1e-4)
+        if cpu_features is None:
+            cpu_features = features
+
+        features = pitch(wave.to(device))
+        assert features.device == device
+        assert torch.allclose(features.cpu(), gt, rtol=1e-4)
+
+if __name__ == "__main__":
+    test_pitch_default()