Merge pull request #235 from kono-dada/fix/inplace-model-load

- Avoid removing and re-inserting model entries during load
- Preserve metadata (bytes, style_vectors) when evicting
- Ensure eviction targets a different loaded model, not always the first
- Reduce unnecessary memory allocations and keep list order stable

Cargo.lock

@@ -1740,7 +1740,7 @@ checksum = "04744f49eae99ab78e0d5c0b603ab218f515ea8cfe5a456d7629ad883a3b6e7d"
 [[package]]
 name = "ort"
 version = "2.0.0-rc.10"
-source = "git+https://github.com/pykeio/ort.git#d28c835c3cc98bcbefc208dc26c8618ccbadec3f"
+source = "git+https://github.com/pykeio/ort.git#5f96a2d5857c3fe9f06282dbf4bdcddbca6c5fe6"
 dependencies = [
  "libloading",
  "ndarray",
@@ -1752,7 +1752,7 @@ dependencies = [
 [[package]]
 name = "ort-sys"
 version = "2.0.0-rc.10"
-source = "git+https://github.com/pykeio/ort.git#d28c835c3cc98bcbefc208dc26c8618ccbadec3f"
+source = "git+https://github.com/pykeio/ort.git#5f96a2d5857c3fe9f06282dbf4bdcddbca6c5fe6"
 dependencies = [
  "flate2",
  "pkg-config",
@@ -2430,9 +2430,9 @@ dependencies = [
 
 [[package]]
 name = "serde_json"
-version = "1.0.141"
+version = "1.0.142"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "30b9eff21ebe718216c6ec64e1d9ac57087aad11efc64e32002bce4a0d4c03d3"
+checksum = "030fedb782600dcbd6f02d479bf0d817ac3bb40d644745b769d6a96bc3afc5a7"
 dependencies = [
  "itoa",
  "memchr",
@@ -2766,9 +2766,9 @@ checksum = "1f3ccbac311fea05f86f61904b462b55fb3df8837a366dfc601a0161d0532f20"
 
 [[package]]
 name = "tokenizers"
-version = "0.21.2"
+version = "0.21.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "4c3846d8588abed0daba25a0e47edd58ea15e450a6088b2575f5116fdb0b27ca"
+checksum = "a620b996116a59e184c2fa2dfd8251ea34a36d0a514758c6f966386bd2e03476"
 dependencies = [
  "ahash",
  "aho-corasick",
@@ -2801,9 +2801,9 @@ dependencies = [
 
 [[package]]
 name = "tokio"
-version = "1.47.0"
+version = "1.47.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "43864ed400b6043a4757a25c7a64a8efde741aed79a056a2fb348a406701bb35"
+checksum = "89e49afdadebb872d3145a5638b59eb0691ea23e46ca484037cfab3b76b95038"
 dependencies = [
  "backtrace",
  "bytes",

crates/sbv2_api/Cargo.toml

@@ -16,7 +16,7 @@ env_logger.workspace = true
 log = "0.4.22"
 sbv2_core = { version = "0.2.0-alpha6", path = "../sbv2_core", features = ["aivmx"] }
 serde = { version = "1.0.210", features = ["derive"] }
-tokio = { version = "1.47.0", features = ["full"] }
+tokio = { version = "1.47.1", features = ["full"] }
 utoipa = { version = "5.4.0", features = ["axum_extras"] }
 utoipa-scalar = { version = "0.3.0", features = ["axum"] }
 

crates/sbv2_core/Cargo.toml

@@ -22,10 +22,10 @@ once_cell.workspace = true
 ort = { git = "https://github.com/pykeio/ort.git", version = "2.0.0-rc.9", optional = true }
 regex = "1.10.6"
 serde = { version = "1.0.210", features = ["derive"] }
-serde_json = "1.0.141"
+serde_json = "1.0.142"
 tar = "0.4.41"
 thiserror = "2.0.11"
-tokenizers = { version = "0.21.2", default-features = false }
+tokenizers = { version = "0.21.4", default-features = false }
 zstd = "0.13.2"
 
 [features]

crates/sbv2_core/src/jtalk.rs

@@ -127,20 +127,20 @@ impl JTalkProcess {
             Ok(phone_tone_list)
         } else if tone_values.len() == 2 {
             if tone_values == hash_set![0, 1] {
-                return Ok(phone_tone_list);
+                Ok(phone_tone_list)
             } else if tone_values == hash_set![-1, 0] {
-                return Ok(phone_tone_list
+                Ok(phone_tone_list
                     .iter()
                     .map(|x| {
                         let new_tone = if x.1 == -1 { 0 } else { 1 };
                         (x.0.clone(), new_tone)
                     })
-                    .collect());
+                    .collect())
             } else {
-                return Err(Error::ValueError("Invalid tone values 0".to_string()));
+                Err(Error::ValueError("Invalid tone values 0".to_string()))
             }
         } else {
-            return Err(Error::ValueError("Invalid tone values 1".to_string()));
+            Err(Error::ValueError("Invalid tone values 1".to_string()))
         }
     }
 

crates/sbv2_core/src/tts.rs

@@ -240,39 +240,43 @@ impl TTSModelHolder {
     }
     fn find_and_load_model<I: Into<TTSIdent>>(&mut self, ident: I) -> Result<bool> {
         let ident = ident.into();
-        let (bytes, style_vectors) = {
-            let model = self
+        // Locate target model entry
+        let target_index = self
             .models
             .iter()
-                .find(|m| m.ident == ident)
+            .position(|m| m.ident == ident)
             .ok_or(Error::ModelNotFoundError(ident.to_string()))?;
-            if model.vits2.is_some() {
+
+        // Already loaded
+        if self.models[target_index].vits2.is_some() {
             return Ok(true);
         }
-            (model.bytes.clone().unwrap(), model.style_vectors.clone())
-        };
-        self.unload(ident.clone());
-        let s = model::load_model(&bytes, false)?;
+
+        // Get bytes to build a Session
+        let bytes = self.models[target_index]
+            .bytes
+            .clone()
+            .ok_or(Error::ModelNotFoundError(ident.to_string()))?;
+
+        // Enforce max loaded models by evicting a different loaded model's session, not removing the entry
         if let Some(max) = self.max_loaded_models {
-            if self.models.iter().filter(|x| x.vits2.is_some()).count() >= max {
-                self.unload(self.models.first().unwrap().ident.clone());
-            }
-        }
-        self.models.push(TTSModel {
-            bytes: Some(bytes.to_vec()),
-            vits2: Some(s),
-            style_vectors,
-            ident: ident.clone(),
-        });
-        let model = self
+            let loaded_count = self.models.iter().filter(|m| m.vits2.is_some()).count();
+            if loaded_count >= max {
+                if let Some(evict_index) = self
                     .models
                     .iter()
-            .find(|m| m.ident == ident)
-            .ok_or(Error::ModelNotFoundError(ident.to_string()))?;
-        if model.vits2.is_some() {
-            return Ok(true);
+                    .position(|m| m.vits2.is_some() && m.ident != ident)
+                {
+                    // Drop only the session to free memory; keep bytes/style for future reload
+                    self.models[evict_index].vits2 = None;
                 }
-        Err(Error::ModelNotFoundError(ident.to_string()))
+            }
+        }
+
+        // Build and set session in-place for the target model
+        let s = model::load_model(&bytes, false)?;
+        self.models[target_index].vits2 = Some(s);
+        Ok(true)
     }
 
     /// Get style vector by style id and weight

crates/sbv2_core/src/tts_util.rs

@@ -173,8 +173,15 @@ pub fn parse_text_blocking(
 }
 
 pub fn array_to_vec(audio_array: Array3<f32>) -> Result<Vec<u8>> {
+    // If SBV2_FORCE_STEREO is set ("1"/"true"), duplicate mono to stereo
+    let force_stereo = std::env::var("SBV2_FORCE_STEREO")
+        .ok()
+        .map(|v| matches!(v.as_str(), "1" | "true" | "TRUE" | "True"))
+        .unwrap_or(false);
+
+    let channels: u16 = if force_stereo { 2 } else { 1 };
     let spec = WavSpec {
-        channels: 1,
+        channels,
         sample_rate: 44100,
         bits_per_sample: 32,
         sample_format: SampleFormat::Float,
@@ -183,10 +190,18 @@ pub fn array_to_vec(audio_array: Array3<f32>) -> Result<Vec<u8>> {
     let mut writer = WavWriter::new(&mut cursor, spec)?;
     for i in 0..audio_array.shape()[0] {
         let output = audio_array.slice(s![i, 0, ..]).to_vec();
+        if force_stereo {
+            for sample in output {
+                // Write to Left and Right channels
+                writer.write_sample(sample)?;
+                writer.write_sample(sample)?;
+            }
+        } else {
             for sample in output {
                 writer.write_sample(sample)?;
             }
         }
+    }
     writer.finalize()?;
     Ok(cursor.into_inner())
 }