chore: sync with upstream

Author: iChizer0

components/acoustics-porting/porting/algorithm/node/speech_commands_preprocess.hpp

@@ -20,7 +20,7 @@
 #include <string>
 #include <string_view>
 
-namespace porting::algorithms::node {
+namespace porting::algorithm::node {
 
 class SpeechCommandsPreprocess final: public module::MNode
 {
@@ -275,7 +275,7 @@ class SpeechCommandsPreprocess final: public module::MNode
     float *_window_buffer;
 };
 
-} // namespace porting::algorithms::node
+} // namespace porting::algorithm::node
 
 #endif // SPEECH_COMMANDS_PREPROCESS_HPP
 

components/acoustics/api/v1/task_sc.hpp

@@ -43,7 +43,7 @@ namespace shared {
     inline std::atomic<bool> is_sampling = false;
     inline std::atomic<bool> is_invoking = false;
 
-    inline constexpr const size_t buffer_size = 102400;
+    inline constexpr const size_t buffer_size = 131072;
     inline constexpr const size_t buffer_size_mask = buffer_size - 1;
     inline std::mutex buffer_mutex;
     inline size_t buffer_head = 0;
@@ -188,7 +188,8 @@ struct TaskSC final
                     const std::lock_guard<std::mutex> lock(v1::shared::buffer_mutex);
                     const size_t head = v1::shared::buffer_head;
                     const size_t tail = v1::shared::buffer_tail;
-                    const size_t next_head = head + shape[0];
+                    const size_t next_head = head + size;
+
                     if (static_cast<size_t>(next_head - tail) > v1::shared::buffer_size) [[unlikely]]
                     {
                         v1::shared::buffer_tail = next_head - v1::shared::buffer_size;

scripts/speech_commands_model_converter.py

@@ -97,7 +97,7 @@ def _get_concrete_input_shape(model: tf.Module) -> Optional[tuple[int, ...]]:
 def convert(
     tfjs_dir: Path,
     output_tflite: Path,
-    quantize: str = "none",
+    quantize: str,
     rep_data: Optional[Path] = None,
     input_shape_override: Optional[str] = None,
 ) -> Path: