simplify

Mobile-Artificial-Intelligence · Aug 13, 2024 · db9e393 · db9e393
1 parent e2774c8
commit db9e393
Show file tree

Hide file tree

Showing 3 changed files with 22 additions and 22 deletions.
diff --git a/include/babylon.hpp b/include/babylon.hpp
@@ -38,9 +38,6 @@ namespace DeepPhonemizer {
       std::vector<std::string> g2p(const std::string& text);
 
     private:
-      const std::array<const char *, 1> input_names = {"text"};
-      const std::array<const char *, 1> output_names = {"output"};
-
       std::string lang;
       bool punctuation;
       Ort::Session* session;
@@ -72,9 +69,6 @@ namespace Vits {
       void tts(const std::vector<std::string>& phonemes, const std::string& output_path);
 
     private:
-      const std::array<const char *, 3> input_names = {"input", "input_lengths", "scales"};
-      const std::array<const char *, 1> output_names = {"output"};
-
       int sample_rate;
       std::vector<float> scales;
 

diff --git a/src/phonemizer.cpp b/src/phonemizer.cpp
@@ -5,6 +5,25 @@
 #include <algorithm>
 #include <cmath>
 
+const std::array<const char *, 1> input_names = {"text"};
+const std::array<const char *, 1> output_names = {"output"};
+
+std::vector<float> softmax(const std::vector<float>& logits) {
+    float max_logit = *std::max_element(logits.begin(), logits.end());
+    std::vector<float> probabilities(logits.size());
+
+    float sum = 0.0f;
+    for (float logit : logits) {
+        sum += std::exp(logit - max_logit);
+    }
+
+    for (size_t i = 0; i < logits.size(); ++i) {
+        probabilities[i] = std::exp(logits[i] - max_logit) / sum;
+    }
+
+    return probabilities;
+}
+
 namespace DeepPhonemizer {
     SequenceTokenizer::SequenceTokenizer(const std::vector<std::string>& symbols, const std::vector<std::string>& languages, int char_repeats, bool lowercase, bool append_start_end)
         : char_repeats(char_repeats), lowercase(lowercase), append_start_end(append_start_end), pad_token("_"), end_token("<end>") {
@@ -113,22 +132,6 @@ namespace DeepPhonemizer {
         return "<" + language + ">";
     }
 
-    std::vector<float> softmax(const std::vector<float>& logits) {
-        float max_logit = *std::max_element(logits.begin(), logits.end());
-        std::vector<float> probabilities(logits.size());
-
-        float sum = 0.0f;
-        for (float logit : logits) {
-            sum += std::exp(logit - max_logit);
-        }
-
-        for (size_t i = 0; i < logits.size(); ++i) {
-            probabilities[i] = std::exp(logits[i] - max_logit) / sum;
-        }
-
-        return probabilities;
-    }
-
     Session::Session(const std::string& model_path, const std::string language, const bool use_punctuation) {
         Ort::Env env(ORT_LOGGING_LEVEL_WARNING, "DeepPhonemizer");
         env.DisableTelemetryEvents();

diff --git a/src/voice.cpp b/src/voice.cpp
@@ -7,6 +7,9 @@
 #include <algorithm>
 #include <cmath>
 
+const std::array<const char *, 3> input_names = {"input", "input_lengths", "scales"};
+const std::array<const char *, 1> output_names = {"output"};
+
 struct WavHeader {
   uint8_t RIFF[4] = {'R', 'I', 'F', 'F'};
   uint32_t chunk_size;