Change how options are displayed (#62)

* Change the way how FrameExtractionOptions is displayed

* small fixes

* Change the way how MelBankOptions is displayed

* Change the way how SpectrogramOptions is displayed

* Change the way how FbankOptions is displayed

* Change the way how MfccOptions is displayed

* Change the way how PlpOptions is displayed

* Export num_frames

* release v1.22

CMakeLists.txt

@@ -7,7 +7,7 @@ project(kaldifeat)
 # remember to change the version in
 # scripts/conda/kaldifeat/meta.yaml
 # scripts/conda-cpu/kaldifeat/meta.yaml
-set(kaldifeat_VERSION "1.21")
+set(kaldifeat_VERSION "1.22")
 
 set(CMAKE_ARCHIVE_OUTPUT_DIRECTORY "${CMAKE_BINARY_DIR}/lib")
 set(CMAKE_LIBRARY_OUTPUT_DIRECTORY "${CMAKE_BINARY_DIR}/lib")

kaldifeat/csrc/feature-fbank.h

@@ -44,20 +44,18 @@ struct FbankOptions {
 
   std::string ToString() const {
     std::ostringstream os;
-    os << "frame_opts: \n";
-    os << frame_opts << "\n";
-    os << "\n";
-
-    os << "mel_opts: \n";
-    os << mel_opts << "\n";
-
-    os << "use_energy: " << use_energy << "\n";
-    os << "energy_floor: " << energy_floor << "\n";
-    os << "raw_energy: " << raw_energy << "\n";
-    os << "htk_compat: " << htk_compat << "\n";
-    os << "use_log_fbank: " << use_log_fbank << "\n";
-    os << "use_power: " << use_power << "\n";
-    os << "device: " << device << "\n";
+    os << "FbankOptions(";
+
+    os << "frame_opts=" << frame_opts.ToString() << ", ";
+    os << "mel_opts=" << mel_opts.ToString() << ", ";
+
+    os << "use_energy=" << (use_energy ? "True" : "False") << ", ";
+    os << "energy_floor=" << energy_floor << ", ";
+    os << "raw_energy=" << (raw_energy ? "True" : "False") << ", ";
+    os << "htk_compat=" << (htk_compat ? "True" : "False") << ", ";
+    os << "use_log_fbank=" << (use_log_fbank ? "True" : "False") << ", ";
+    os << "use_power=" << (use_power ? "True" : "False") << ", ";
+    os << "device=\"" << device << "\")";
     return os.str();
   }
 };

kaldifeat/csrc/feature-mfcc.h

@@ -53,20 +53,18 @@ struct MfccOptions {
 
   std::string ToString() const {
     std::ostringstream os;
-    os << "frame_opts: \n";
-    os << frame_opts << "\n";
-    os << "\n";
-
-    os << "mel_opts: \n";
-    os << mel_opts << "\n";
-
-    os << "num_ceps: " << num_ceps << "\n";
-    os << "use_energy: " << use_energy << "\n";
-    os << "energy_floor: " << energy_floor << "\n";
-    os << "raw_energy: " << raw_energy << "\n";
-    os << "cepstral_lifter: " << cepstral_lifter << "\n";
-    os << "htk_compat: " << htk_compat << "\n";
-    os << "device: " << device << "\n";
+    os << "MfccOptions(";
+    os << "frame_opts=" << frame_opts.ToString() << ", ";
+    os << "mel_opts=" << mel_opts.ToString() << ", ";
+
+    os << "num_ceps=" << num_ceps << ", ";
+    os << "use_energy=" << (use_energy ? "True" : "False") << ", ";
+    os << "energy_floor=" << energy_floor << ", ";
+    os << "raw_energy=" << (raw_energy ? "True" : "False") << ", ";
+    os << "cepstral_lifter=" << cepstral_lifter << ", ";
+    os << "htk_compat=" << (htk_compat ? "True" : "False") << ", ";
+    os << "device=\"" << device << "\")";
+
     return os.str();
   }
 };

kaldifeat/csrc/feature-plp.h

@@ -61,23 +61,21 @@ struct PlpOptions {
 
   std::string ToString() const {
     std::ostringstream os;
-    os << "frame_opts: \n";
-    os << frame_opts << "\n";
-    os << "\n";
-
-    os << "mel_opts: \n";
-    os << mel_opts << "\n";
-
-    os << "lpc_order: " << lpc_order << "\n";
-    os << "num_ceps: " << num_ceps << "\n";
-    os << "use_energy: " << use_energy << "\n";
-    os << "energy_floor: " << energy_floor << "\n";
-    os << "raw_energy: " << raw_energy << "\n";
-    os << "compress_factor: " << compress_factor << "\n";
-    os << "cepstral_lifter: " << cepstral_lifter << "\n";
-    os << "cepstral_scale: " << cepstral_scale << "\n";
-    os << "htk_compat: " << htk_compat << "\n";
-    os << "device: " << device << "\n";
+    os << "PlpOptions(";
+
+    os << "frame_opts=" << frame_opts.ToString() << ", ";
+    os << "mel_opts=" << mel_opts.ToString() << ", ";
+
+    os << "lpc_order=" << lpc_order << ", ";
+    os << "num_ceps=" << num_ceps << ", ";
+    os << "use_energy=" << (use_energy ? "True" : "False") << ", ";
+    os << "energy_floor=" << energy_floor << ", ";
+    os << "raw_energy=" << (raw_energy ? "True" : "False") << ", ";
+    os << "compress_factor=" << compress_factor << ", ";
+    os << "cepstral_lifter=" << cepstral_lifter << ", ";
+    os << "cepstral_scale=" << cepstral_scale << ", ";
+    os << "htk_compat=" << (htk_compat ? "True" : "False") << ", ";
+    os << "device=\"" << device << "\")";
     return os.str();
   }
 };

kaldifeat/csrc/feature-spectrogram.h

@@ -36,13 +36,12 @@ struct SpectrogramOptions {
 
   std::string ToString() const {
     std::ostringstream os;
-    os << "frame_opts: \n";
-    os << frame_opts << "\n";
-
-    os << "energy_floor: " << energy_floor << "\n";
-    os << "raw_energy: " << raw_energy << "\n";
-    // os << "return_raw_fft: " << return_raw_fft << "\n";
-    os << "device: " << device << "\n";
+    os << "SpectrogramOptions(";
+    os << "frame_opts=" << frame_opts.ToString() << ", ";
+    os << "energy_floor=" << energy_floor << ", ";
+    os << "raw_energy=" << (raw_energy ? "True" : "False") << ", ";
+    os << "return_raw_fft=" << (return_raw_fft ? "True" : "False") << ", ";
+    os << "device=\"" << device << "\")";
     return os.str();
   }
 };

kaldifeat/csrc/feature-window.h

@@ -62,21 +62,20 @@ struct FrameExtractionOptions {
   }
   std::string ToString() const {
     std::ostringstream os;
-#define KALDIFEAT_PRINT(x) os << #x << ": " << x << "\n"
-    KALDIFEAT_PRINT(samp_freq);
-    KALDIFEAT_PRINT(frame_shift_ms);
-    KALDIFEAT_PRINT(frame_length_ms);
-    KALDIFEAT_PRINT(dither);
-    KALDIFEAT_PRINT(preemph_coeff);
-    KALDIFEAT_PRINT(remove_dc_offset);
-    KALDIFEAT_PRINT(window_type);
-    KALDIFEAT_PRINT(round_to_power_of_two);
-    KALDIFEAT_PRINT(blackman_coeff);
-    KALDIFEAT_PRINT(snip_edges);
-    // KALDIFEAT_PRINT(allow_downsample);
-    // KALDIFEAT_PRINT(allow_upsample);
-    KALDIFEAT_PRINT(max_feature_vectors);
-#undef KALDIFEAT_PRINT
+    os << "FrameExtractionOptions(";
+    os << "samp_freq=" << samp_freq << ", ";
+    os << "frame_shift_ms=" << frame_shift_ms << ", ";
+    os << "frame_length_ms=" << frame_length_ms << ", ";
+    os << "dither=" << dither << ", ";
+    os << "preemph_coeff=" << preemph_coeff << ", ";
+    os << "remove_dc_offset=" << (remove_dc_offset ? "True" : "False") << ", ";
+    os << "window_type=" << '"' << window_type << '"' << ", ";
+    os << "round_to_power_of_two=" << (round_to_power_of_two ? "True" : "False")
+       << ", ";
+    os << "blackman_coeff=" << blackman_coeff << ", ";
+    os << "snip_edges=" << (snip_edges ? "True" : "False") << ", ";
+    os << "max_feature_vectors=" << max_feature_vectors << ")";
+
     return os.str();
   }
 };

kaldifeat/csrc/mel-computations.h

@@ -36,13 +36,14 @@ struct MelBanksOptions {
 
   std::string ToString() const {
     std::ostringstream os;
-    os << "num_bins: " << num_bins << "\n";
-    os << "low_freq: " << low_freq << "\n";
-    os << "high_freq: " << high_freq << "\n";
-    os << "vtln_low: " << vtln_low << "\n";
-    os << "vtln_high: " << vtln_high << "\n";
-    os << "debug_mel: " << debug_mel << "\n";
-    os << "htk_mode: " << htk_mode << "\n";
+    os << "MelBanksOptions(";
+    os << "num_bins=" << num_bins << ", ";
+    os << "low_freq=" << low_freq << ", ";
+    os << "high_freq=" << high_freq << ", ";
+    os << "vtln_low=" << vtln_low << ", ";
+    os << "vtln_high=" << vtln_high << ", ";
+    os << "debug_mel=" << (debug_mel ? "True" : "False") << ", ";
+    os << "htk_mode=" << (htk_mode ? "True" : "False") << ")";
     return os.str();
   }
 };

kaldifeat/python/csrc/feature-fbank.cc

@@ -16,6 +16,35 @@ static void PybindFbankOptions(py::module &m) {
   using PyClass = FbankOptions;
   py::class_<PyClass>(m, "FbankOptions")
       .def(py::init<>())
+      .def(py::init([](const MelBanksOptions &mel_opts,
+                       const FrameExtractionOptions &frame_opts =
+                           FrameExtractionOptions(),
+                       bool use_energy = false, float energy_floor = 0.0f,
+                       bool raw_energy = true, bool htk_compat = false,
+                       bool use_log_fbank = true, bool use_power = true,
+                       py::object device =
+                           py::str("cpu")) -> std::unique_ptr<FbankOptions> {
+             auto opts = std::make_unique<FbankOptions>();
+             opts->frame_opts = frame_opts;
+             opts->mel_opts = mel_opts;
+             opts->use_energy = use_energy;
+             opts->energy_floor = energy_floor;
+             opts->raw_energy = raw_energy;
+             opts->htk_compat = htk_compat;
+             opts->use_log_fbank = use_log_fbank;
+             opts->use_power = use_power;
+
+             std::string s = static_cast<py::str>(device);
+             opts->device = torch::Device(s);
+
+             return opts;
+           }),
+           py::arg("mel_opts"),
+           py::arg("frame_opts") = FrameExtractionOptions(),
+           py::arg("use_energy") = false, py::arg("energy_floor") = 0.0f,
+           py::arg("raw_energy") = true, py::arg("htk_compat") = false,
+           py::arg("use_log_fbank") = true, py::arg("use_power") = true,
+           py::arg("device") = py::str("cpu"))
       .def_readwrite("frame_opts", &PyClass::frame_opts)
       .def_readwrite("mel_opts", &PyClass::mel_opts)
       .def_readwrite("use_energy", &PyClass::use_energy)

kaldifeat/python/csrc/feature-mfcc.cc

@@ -16,6 +16,35 @@ void PybindMfccOptions(py::module &m) {
   using PyClass = MfccOptions;
   py::class_<PyClass>(m, "MfccOptions")
       .def(py::init<>())
+      .def(py::init([](const MelBanksOptions &mel_opts,
+                       const FrameExtractionOptions &frame_opts =
+                           FrameExtractionOptions(),
+                       int32_t num_ceps = 13, bool use_energy = true,
+                       float energy_floor = 0.0, bool raw_energy = true,
+                       float cepstral_lifter = 22.0, bool htk_compat = false,
+                       py::object device =
+                           py::str("cpu")) -> std::unique_ptr<MfccOptions> {
+             auto opts = std::make_unique<MfccOptions>();
+             opts->frame_opts = frame_opts;
+             opts->mel_opts = mel_opts;
+             opts->num_ceps = num_ceps;
+             opts->use_energy = use_energy;
+             opts->energy_floor = energy_floor;
+             opts->raw_energy = raw_energy;
+             opts->cepstral_lifter = cepstral_lifter;
+             opts->htk_compat = htk_compat;
+
+             std::string s = static_cast<py::str>(device);
+             opts->device = torch::Device(s);
+
+             return opts;
+           }),
+           py::arg("mel_opts"),
+           py::arg("frame_opts") = FrameExtractionOptions(),
+           py::arg("num_ceps") = 13, py::arg("use_energy") = true,
+           py::arg("energy_floor") = 0.0f, py::arg("raw_energy") = true,
+           py::arg("cepstral_lifter") = 22.0, py::arg("htk_compat") = false,
+           py::arg("device") = py::str("cpu"))
       .def_readwrite("frame_opts", &PyClass::frame_opts)
       .def_readwrite("mel_opts", &PyClass::mel_opts)
       .def_readwrite("num_ceps", &PyClass::num_ceps)

kaldifeat/python/csrc/feature-plp.cc

@@ -16,6 +16,41 @@ void PybindPlpOptions(py::module &m) {
   using PyClass = PlpOptions;
   py::class_<PyClass>(m, "PlpOptions")
       .def(py::init<>())
+      .def(py::init([](const MelBanksOptions &mel_opts,
+                       const FrameExtractionOptions &frame_opts =
+                           FrameExtractionOptions(),
+                       int32_t lpc_order = 12, int32_t num_ceps = 13,
+                       bool use_energy = true, float energy_floor = 0.0,
+                       bool raw_energy = true, float compress_factor = 0.33333,
+                       int32_t cepstral_lifter = 22, float cepstral_scale = 1.0,
+                       bool htk_compat = false,
+                       py::object device =
+                           py::str("cpu")) -> std::unique_ptr<PlpOptions> {
+             auto opts = std::make_unique<PlpOptions>();
+             opts->frame_opts = frame_opts;
+             opts->mel_opts = mel_opts;
+             opts->lpc_order = lpc_order;
+             opts->num_ceps = num_ceps;
+             opts->use_energy = use_energy;
+             opts->energy_floor = energy_floor;
+             opts->raw_energy = raw_energy;
+             opts->compress_factor = compress_factor;
+             opts->cepstral_lifter = cepstral_lifter;
+             opts->cepstral_scale = cepstral_scale;
+             opts->htk_compat = htk_compat;
+
+             std::string s = static_cast<py::str>(device);
+             opts->device = torch::Device(s);
+
+             return opts;
+           }),
+           py::arg("mel_opts"),
+           py::arg("frame_opts") = FrameExtractionOptions(),
+           py::arg("lpc_order") = 12, py::arg("num_ceps") = 13,
+           py::arg("use_energy") = true, py::arg("energy_floor") = 0.0,
+           py::arg("raw_energy") = true, py::arg("compress_factor") = 0.33333,
+           py::arg("cepstral_lifter") = 22, py::arg("cepstral_scale") = 1.0,
+           py::arg("htk_compat") = false, py::arg("device") = py::str("cpu"))
       .def_readwrite("frame_opts", &PyClass::frame_opts)
       .def_readwrite("mel_opts", &PyClass::mel_opts)
       .def_readwrite("lpc_order", &PyClass::lpc_order)

kaldifeat/python/csrc/feature-spectrogram.cc

@@ -15,7 +15,27 @@ namespace kaldifeat {
 static void PybindSpectrogramOptions(py::module &m) {
   using PyClass = SpectrogramOptions;
   py::class_<PyClass>(m, "SpectrogramOptions")
-      .def(py::init<>())
+      .def(py::init([](const FrameExtractionOptions &frame_opts =
+                           FrameExtractionOptions(),
+                       float energy_floor = 0.0, bool raw_energy = true,
+                       bool return_raw_fft = false,
+                       py::object device = py::str(
+                           "cpu")) -> std::unique_ptr<SpectrogramOptions> {
+             auto opts = std::make_unique<SpectrogramOptions>();
+             opts->frame_opts = frame_opts;
+             opts->energy_floor = energy_floor;
+             opts->raw_energy = raw_energy;
+             opts->return_raw_fft = return_raw_fft;
+
+             std::string s = static_cast<py::str>(device);
+             opts->device = torch::Device(s);
+
+             return opts;
+           }),
+           py::arg("frame_opts") = FrameExtractionOptions(),
+           py::arg("energy_floor") = 0.0, py::arg("raw_energy") = true,
+           py::arg("return_raw_fft") = false,
+           py::arg("device") = py::str("cpu"))
       .def_readwrite("frame_opts", &PyClass::frame_opts)
       .def_readwrite("energy_floor", &PyClass::energy_floor)
       .def_readwrite("raw_energy", &PyClass::raw_energy)

kaldifeat/python/csrc/feature-window.cc

@@ -4,6 +4,7 @@
 
 #include "kaldifeat/python/csrc/feature-window.h"
 
+#include <memory>
 #include <string>
 
 #include "kaldifeat/csrc/feature-window.h"
@@ -14,7 +15,38 @@ namespace kaldifeat {
 static void PybindFrameExtractionOptions(py::module &m) {
   using PyClass = FrameExtractionOptions;
   py::class_<PyClass>(m, "FrameExtractionOptions")
-      .def(py::init<>())
+      .def(
+          py::init([](float samp_freq = 16000, float frame_shift_ms = 10.0f,
+                      float frame_length_ms = 25.0f, float dither = 1.0f,
+                      float preemph_coeff = 0.97f, bool remove_dc_offset = true,
+                      const std::string &window_type = "povey",
+                      bool round_to_power_of_two = true,
+                      float blackman_coeff = 0.42f, bool snip_edges = true,
+                      int32_t max_feature_vectors =
+                          -1) -> std::unique_ptr<FrameExtractionOptions> {
+            auto opts = std::make_unique<FrameExtractionOptions>();
+
+            opts->samp_freq = samp_freq;
+            opts->frame_shift_ms = frame_shift_ms;
+            opts->frame_length_ms = frame_length_ms;
+            opts->dither = dither;
+            opts->preemph_coeff = preemph_coeff;
+            opts->remove_dc_offset = remove_dc_offset;
+            opts->window_type = window_type;
+            opts->round_to_power_of_two = round_to_power_of_two;
+            opts->blackman_coeff = blackman_coeff;
+            opts->snip_edges = snip_edges;
+            opts->max_feature_vectors = max_feature_vectors;
+
+            return opts;
+          }),
+          py::arg("samp_freq") = 16000, py::arg("frame_shift_ms") = 10.0f,
+          py::arg("frame_length_ms") = 25.0f, py::arg("dither") = 1.0f,
+          py::arg("preemph_coeff") = 0.97f, py::arg("remove_dc_offset") = true,
+          py::arg("window_type") = "povey",
+          py::arg("round_to_power_of_two") = true,
+          py::arg("blackman_coeff") = 0.42f, py::arg("snip_edges") = true,
+          py::arg("max_feature_vectors") = -1)
       .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)