Clean code a bit

Author: nenad1002

CMakeLists.txt

@@ -778,7 +778,6 @@ target_link_libraries(ocos_operators PRIVATE ${ocos_libraries})
 
 file(GLOB _TARGET_LIB_SRC "shared/lib/*.cc")
 
-# NeMo mel spectrogram is standalone (no ORT/C API deps) — always compile when audio is enabled
 if(OCOS_ENABLE_AUDIO)
   file(GLOB nemo_mel_SRC "shared/api/nemo_mel_*")
   list(APPEND _TARGET_LIB_SRC ${nemo_mel_SRC})

shared/api/nemo_mel_spectrogram.cc

@@ -20,15 +20,13 @@
 
 namespace nemo_mel {
 
-// ─── Slaney mel scale constants ─────────────────────────────────────────────
+// Slaney mel scale constants
 
 static constexpr float kMinLogHz = 1000.0f;
 static constexpr float kMinLogMel = 15.0f;               // 1000 / (200/3)
 static constexpr float kLinScale = 200.0f / 3.0f;        // Hz per mel (linear region)
 static constexpr float kLogStep = 0.06875177742094912f;   // log(6.4) / 27
 
-// ─── Mel scale conversions ──────────────────────────────────────────────────
-
 float HzToMel(float hz) {
   if (hz < kMinLogHz) return hz / kLinScale;
   return kMinLogMel + std::log(hz / kMinLogHz) / kLogStep;
@@ -39,8 +37,6 @@ float MelToHz(float mel) {
   return kMinLogHz * std::exp((mel - kMinLogMel) * kLogStep);
 }
 
-// ─── Filterbank creation ────────────────────────────────────────────────────
-
 std::vector<std::vector<float>> CreateMelFilterbank(int num_mels, int fft_size, int sample_rate) {
   int num_bins = fft_size / 2 + 1;
   float mel_low = HzToMel(0.0f);
@@ -82,8 +78,6 @@ std::vector<std::vector<float>> CreateMelFilterbank(int num_mels, int fft_size,
   return filterbank;
 }
 
-// ─── Single-frame DFT ──────────────────────────────────────────────────────
-
 void ComputeSTFTFrame(const float* frame, const float* window, int frame_len,
                       int fft_size, std::vector<float>& magnitudes) {
   int num_bins = fft_size / 2 + 1;
@@ -107,8 +101,7 @@ void ComputeSTFTFrame(const float* frame, const float* window, int frame_len,
   }
 }
 
-// ─── Batch (offline) log-mel extraction ─────────────────────────────────────
-
+// BATCH LOG-MEL EXTRACTION
 std::vector<float> NemoComputeLogMelBatch(const float* audio, size_t num_samples,
                                           const NemoMelConfig& cfg, int& out_num_frames) {
   // Lazily-initialized statics are fine for batch mode (same config per process).
@@ -163,8 +156,7 @@ std::vector<float> NemoComputeLogMelBatch(const float* audio, size_t num_samples
   return mel_spec;
 }
 
-// ─── Streaming log-mel extraction ───────────────────────────────────────────
-
+// STREAMING LOG-MEL EXTRACTION
 NemoStreamingMelExtractor::NemoStreamingMelExtractor(const NemoMelConfig& cfg)
     : cfg_(cfg) {
   mel_filters_ = CreateMelFilterbank(cfg_.num_mels, cfg_.fft_size, cfg_.sample_rate);
@@ -192,7 +184,7 @@ std::pair<std::vector<float>, int> NemoStreamingMelExtractor::Process(
 
   // Left-only center pad for streaming: prepend overlap from previous chunk.
   // For the first chunk this is zeros (matching center=True left edge).
-  int pad = cfg_.fft_size / 2;  // 256 samples
+  int pad = cfg_.fft_size / 2;
   std::vector<float> padded(pad + num_samples);
   std::memcpy(padded.data(), audio_overlap_.data(), pad * sizeof(float));
   std::memcpy(padded.data() + pad, preemphasized.data(), num_samples * sizeof(float));

shared/api/nemo_mel_spectrogram.h

@@ -2,8 +2,6 @@
 // Licensed under the MIT License.
 //
 // NeMo-compatible log-mel spectrogram extraction (Slaney scale, matching librosa/NeMo).
-// No ONNX Runtime or other framework dependencies — pure C++ with standard library only.
-// Designed to be portable across repos.
 
 #pragma once
 
@@ -13,8 +11,6 @@
 
 namespace nemo_mel {
 
-// ─── Configuration ──────────────────────────────────────────────────────────
-
 struct NemoMelConfig {
   int num_mels;
   int fft_size;
@@ -25,19 +21,15 @@ struct NemoMelConfig {
   float log_eps;
 };
 
-// ─── Mel scale conversions (Slaney) ────────────────────────────────────────
+// Mel scale conversions (Slaney)
 
 float HzToMel(float hz);
 float MelToHz(float mel);
 
-// ─── Filterbank creation ────────────────────────────────────────────────────
-
 /// Build a triangular mel filterbank with Slaney normalization (matches librosa).
 /// Returns shape [num_mels][num_bins] where num_bins = fft_size/2 + 1.
 std::vector<std::vector<float>> CreateMelFilterbank(int num_mels, int fft_size, int sample_rate);
 
-// ─── Single-frame DFT ──────────────────────────────────────────────────────
-
 /// Compute |DFT|^2 (power spectrum) for a single windowed frame.
 /// frame: pointer to fft_size samples (or win_length samples with window applied).
 /// window: pointer to window coefficients (same length as frame_len).
@@ -47,8 +39,8 @@ std::vector<std::vector<float>> CreateMelFilterbank(int num_mels, int fft_size,
 void ComputeSTFTFrame(const float* frame, const float* window, int frame_len,
                       int fft_size, std::vector<float>& magnitudes);
 
-// ─── Batch (offline) log-mel extraction ─────────────────────────────────────
 
+// BATCH LOG-MEL EXTRACTION
 /// Compute NeMo-compatible log-mel spectrogram for a complete audio buffer.
 /// Applies pre-emphasis, center-pads both sides (fft_size/2 zeros), computes STFT
 /// with a periodic Hann window, applies mel filterbank, and takes log(mel + eps).
@@ -58,8 +50,7 @@ void ComputeSTFTFrame(const float* frame, const float* window, int frame_len,
 std::vector<float> NemoComputeLogMelBatch(const float* audio, size_t num_samples,
                                           const NemoMelConfig& cfg, int& out_num_frames);
 
-// ─── Streaming log-mel extraction ───────────────────────────────────────────
-
+// STREAMING LOG-MEL EXTRACTION
 /// Stateful streaming NeMo-compatible mel extractor that maintains overlap and
 /// pre-emphasis state across successive audio chunks.
 ///

test/static_test/test_nemo_mel.cc

@@ -11,8 +11,6 @@
 
 using namespace nemo_mel;
 
-// ─── Helpers ────────────────────────────────────────────────────────────────
-
 static NemoMelConfig MakeTestConfig() {
   NemoMelConfig cfg;
   cfg.num_mels = 128;
@@ -37,7 +35,7 @@ static std::vector<float> SineWave(float freq_hz, float duration_sec,
   return wav;
 }
 
-// ─── Mel scale conversions ──────────────────────────────────────────────────
+// Mel scale conversions (Slaney)
 
 TEST(NemoMelTest, HzToMelLinearRegion) {
   // Below 1000 Hz the Slaney scale is linear: mel = 3 * hz / 200