add half-precision select to rvc

client/demo/dist/assets/gui_settings/RVC.json

@@ -42,7 +42,7 @@
 
                     "showFeature": true,
                     "showIndex": true,
-
+                    "showHalfPrecision": true,
                     "defaultEnablePyTorch": true
                 }
             },

client/demo/public/assets/gui_settings/RVC.json

@@ -42,7 +42,7 @@
 
                     "showFeature": true,
                     "showIndex": true,
-
+                    "showHalfPrecision": true,
                     "defaultEnablePyTorch": true
                 }
             },

client/demo/src/components/demo/components/301-8_HalfPrescisionRow.tsx

@@ -0,0 +1,31 @@
+import React, { useMemo } from "react"
+import { fileSelector } from "@dannadori/voice-changer-client-js"
+import { useAppState } from "../../../001_provider/001_AppStateProvider"
+
+
+export const HalfPrecisionRow = () => {
+    const appState = useAppState()
+
+    const halfPrecisionSelectRow = useMemo(() => {
+        const onHalfPrecisionChanged = () => {
+            appState.serverSetting.setFileUploadSetting({
+                ...appState.serverSetting.fileUploadSetting,
+                isHalf: !appState.serverSetting.fileUploadSetting.isHalf
+            })
+        }
+
+        return (
+            <div className="body-row split-3-3-4 left-padding-1 guided">
+                <div className="body-item-title left-padding-2">-</div>
+                <div className="body-item-text">
+                    <div></div>
+                </div>
+                <div className="body-button-container">
+                    <input type="checkbox" checked={appState.serverSetting.fileUploadSetting.isHalf} onChange={() => onHalfPrecisionChanged()} /> half-precision
+                </div>
+            </div>
+        )
+    }, [appState.serverSetting.fileUploadSetting, appState.serverSetting.setFileUploadSetting])
+
+    return halfPrecisionSelectRow
+}

client/demo/src/components/demo/components/301_ModelUploaderRow.tsx

@@ -7,6 +7,7 @@ import { CorrespondenceSelectRow } from "./301-4_CorrespondenceSelectRow"
 import { PyTorchClusterSelectRow } from "./301-5_PyTorchClusterSelectRow"
 import { FeatureSelectRow } from "./301-6_FeatureSelectRow"
 import { IndexSelectRow } from "./301-7_IndexSelectRow"
+import { HalfPrecisionRow } from "./301-8_HalfPrescisionRow"
 import { ModelUploadButtonRow } from "./301-9_ModelUploadButtonRow"
 
 export type ModelUploaderRowProps = {
@@ -18,6 +19,7 @@ export type ModelUploaderRowProps = {
 
     showFeature: boolean
     showIndex: boolean
+    showHalfPrecision: boolean
 
     defaultEnablePyTorch: boolean
 }
@@ -47,6 +49,7 @@ export const ModelUploaderRow = (props: ModelUploaderRowProps) => {
                 {props.showConfig ? <ConfigSelectRow /> : <></>}
                 {props.showOnnx ? <ONNXSelectRow /> : <></>}
                 {props.showPyTorch && guiState.showPyTorchModelUpload ? <PyTorchSelectRow /> : <></>}
+                {props.showHalfPrecision ? <HalfPrecisionRow /> : <></>}
                 {props.showCorrespondence ? <CorrespondenceSelectRow /> : <></>}
                 {props.showPyTorchCluster ? <PyTorchClusterSelectRow /> : <></>}
                 {props.showFeature ? <FeatureSelectRow /> : <></>}

client/lib/src/ServerConfigurator.ts

@@ -110,7 +110,7 @@ export class ServerConfigurator {
     }
 
     // !! 注意!! hubertTorchModelは固定値で上書きされるため、設定しても効果ない。
-    loadModel = async (configFilename: string, pyTorchModelFilename: string | null, onnxModelFilename: string | null, clusterTorchModelFilename: string | null, featureFilename: string | null, indexFilename: string | null) => {
+    loadModel = async (configFilename: string, pyTorchModelFilename: string | null, onnxModelFilename: string | null, clusterTorchModelFilename: string | null, featureFilename: string | null, indexFilename: string | null, isHalf: boolean) => {
         const url = this.serverUrl + "/load_model"
         const info = new Promise<ServerInfo>(async (resolve) => {
             const formData = new FormData();
@@ -120,6 +120,7 @@ export class ServerConfigurator {
             formData.append("clusterTorchModelFilename", clusterTorchModelFilename || "-");
             formData.append("featureFilename", featureFilename || "-");
             formData.append("indexFilename", indexFilename || "-");
+            formData.append("isHalf", "" + isHalf);
 
             const request = new Request(url, {
                 method: 'POST',

client/lib/src/VoiceChangerClient.ts

@@ -264,9 +264,10 @@ export class VoiceChangerClient {
         clusterTorchModelFilename: string | null,
         featureFilename: string | null,
         indexFilename: string | null,
+        isHalf: boolean
     ) => {
         // !! 注意!! hubertTorchModelは固定値で上書きされるため、設定しても効果ない。
-        return this.configurator.loadModel(configFilename, pyTorchModelFilename, onnxModelFilename, clusterTorchModelFilename, featureFilename, indexFilename)
+        return this.configurator.loadModel(configFilename, pyTorchModelFilename, onnxModelFilename, clusterTorchModelFilename, featureFilename, indexFilename, isHalf)
     }
 
     //##  Worklet ##//

client/lib/src/hooks/useServerSetting.ts

@@ -20,6 +20,8 @@ export type FileUploadSetting = {
     feature: ModelData | null //RVC
     index: ModelData | null   //RVC
 
+    isHalf: boolean
+
 }
 
 const InitialFileUploadSetting: FileUploadSetting = {
@@ -30,7 +32,9 @@ const InitialFileUploadSetting: FileUploadSetting = {
     hubertTorchModel: null,
 
     feature: null,
-    index: null
+    index: null,
+
+    isHalf: true
 }
 
 export type UseServerSettingProps = {
@@ -234,13 +238,16 @@ export const useServerSetting = (props: UseServerSettingProps): ServerSettingSta
 
             // !! 注意!! hubertTorchModelは固定値で上書きされるため、設定しても効果ない。
             const configFileName = fileUploadSetting.configFile ? fileUploadSetting.configFile.filename || "-" : "-"
+            console.log("IS HALF", fileUploadSetting.isHalf)
             const loadPromise = props.voiceChangerClient.loadModel(
                 configFileName,
                 fileUploadSetting.pyTorchModel?.filename || null,
                 fileUploadSetting.onnxModel?.filename || null,
                 fileUploadSetting.clusterTorchModel?.filename || null,
                 fileUploadSetting.feature?.filename || null,
-                fileUploadSetting.index?.filename || null
+                fileUploadSetting.index?.filename || null,
+                fileUploadSetting.isHalf
+
             )
 
             // サーバでロード中にキャッシュにセーブ
@@ -260,7 +267,8 @@ export const useServerSetting = (props: UseServerSettingProps): ServerSettingSta
                     } : null,
                     index: fileUploadSetting.index ? {
                         data: fileUploadSetting.index.data, filename: fileUploadSetting.index.filename
-                    } : null
+                    } : null,
+                    isHalf: fileUploadSetting.isHalf
                 }
                 setItem(INDEXEDDB_KEY_MODEL_DATA, saveData)
 

server/restapi/MMVC_Rest_Fileuploader.py

@@ -57,6 +57,7 @@ class MMVC_Rest_Fileuploader:
         clusterTorchModelFilename: str = Form(...),
         featureFilename: str = Form(...),
         indexFilename: str = Form(...),
+        isHalf: bool = Form(...),
     ):
         pyTorchModelFilePath = os.path.join(UPLOAD_DIR, pyTorchModelFilename) if pyTorchModelFilename != "-" else None
         onnxModelFilePath = os.path.join(UPLOAD_DIR, onnxModelFilename) if onnxModelFilename != "-" else None
@@ -66,7 +67,8 @@ class MMVC_Rest_Fileuploader:
         indexFilePath = os.path.join(UPLOAD_DIR, indexFilename) if indexFilename != "-" else None
 
         info = self.voiceChangerManager.loadModel(configFilePath, pyTorchModelFilePath, onnxModelFilePath,
-                                                  clusterTorchModelFilePath, featureFilePath, indexFilePath)
+                                                  clusterTorchModelFilePath, featureFilePath, indexFilePath,
+                                                  isHalf)
         json_compatible_item_data = jsonable_encoder(info)
         return JSONResponse(content=json_compatible_item_data)
         # return {"load": f"{configFilePath}, {pyTorchModelFilePath}, {onnxModelFilePath}"}

server/voice_changer/RVC/RVC.py

@@ -52,14 +52,14 @@ class RVCSettings():
     configFile: str = ""
 
     indexRatio: float = 0
-    isHalf: int = 0
+    quality: int = 0
 
     speakers: dict[str, int] = field(
         default_factory=lambda: {}
     )
 
     # ↓mutableな物だけ列挙
-    intData = ["gpu", "dstId", "tran", "predictF0", "extraConvertSize"]
+    intData = ["gpu", "dstId", "tran", "predictF0", "extraConvertSize", "quality"]
     floatData = ["noiceScale", "silentThreshold", "indexRatio"]
     strData = ["framework", "f0Detector"]
 
@@ -76,12 +76,11 @@ class RVC:
         self.params = params
         print("RVC initialization: ", params)
 
-    def loadModel(self, config: str, pyTorch_model_file: str = None, onnx_model_file: str = None, feature_file: str = None, index_file: str = None):
+    def loadModel(self, config: str, pyTorch_model_file: str = None, onnx_model_file: str = None, feature_file: str = None, index_file: str = None, is_half: bool = True):
         self.settings.configFile = config
         self.feature_file = feature_file
         self.index_file = index_file
-
+        self.is_half = is_half
-        print("featurefile", feature_file, index_file)
 
         self.tgt_sr = 40000
         try:
@@ -89,7 +88,7 @@ class RVC:
             models, saved_cfg, task = checkpoint_utils.load_model_ensemble_and_task([hubert_path], suffix="",)
             model = models[0]
             model.eval()
-            if self.settings.isHalf:
+            if self.is_half:
                 model = model.half()
             self.hubert_model = model
 
@@ -105,17 +104,17 @@ class RVC:
         if pyTorch_model_file != None:
             cpt = torch.load(pyTorch_model_file, map_location="cpu")
             self.tgt_sr = cpt["config"][-1]
-            net_g = SynthesizerTrnMs256NSFsid(*cpt["config"], is_half=self.settings.isHalf)
+            net_g = SynthesizerTrnMs256NSFsid(*cpt["config"], is_half=self.is_half)
             net_g.eval()
             net_g.load_state_dict(cpt["weight"], strict=False)
-            if self.settings.isHalf:
+            if self.is_half:
                 net_g = net_g.half()
             self.net_g = net_g
 
         # ONNXモデル生成
         if onnx_model_file != None:
             # self.onnx_session = ModelWrapper(onnx_model_file, is_half=True)
-            self.onnx_session = ModelWrapper(onnx_model_file, is_half=self.settings.isHalf)
+            self.onnx_session = ModelWrapper(onnx_model_file, is_half=self.is_half)
             # input_info = self.onnx_session.get_inputs()
             # for i in input_info:
             #     print("input", i)
@@ -222,7 +221,7 @@ class RVC:
             return np.zeros(convertSize).astype(np.int16)
 
         with torch.no_grad():
-            vc = VC(self.tgt_sr, dev, self.settings.isHalf)
+            vc = VC(self.tgt_sr, dev, self.is_half)
             sid = 0
             times = [0, 0, 0]
             f0_up_key = self.settings.tran
@@ -263,7 +262,7 @@ class RVC:
             return np.zeros(convertSize).astype(np.int16)
 
         with torch.no_grad():
-            vc = VC(self.tgt_sr, dev, self.settings.isHalf)
+            vc = VC(self.tgt_sr, dev, self.is_half)
             sid = 0
             times = [0, 0, 0]
             f0_up_key = self.settings.tran

server/voice_changer/VoiceChanger.py

@@ -77,13 +77,13 @@ class VoiceChanger():
 
         print(f"VoiceChanger Initialized (GPU_NUM:{self.gpu_num}, mps_enabled:{self.mps_enabled})")
 
-    def loadModel(self, config: str, pyTorch_model_file: str = None, onnx_model_file: str = None, clusterTorchModel: str = None, feature_file: str = None, index_file: str = None):
+    def loadModel(self, config: str, pyTorch_model_file: str = None, onnx_model_file: str = None, clusterTorchModel: str = None, feature_file: str = None, index_file: str = None, is_half: bool = True):
         if self.modelType == "MMVCv15" or self.modelType == "MMVCv13":
             return self.voiceChanger.loadModel(config, pyTorch_model_file, onnx_model_file)
         elif self.modelType == "so-vits-svc-40" or self.modelType == "so-vits-svc-40_c" or self.modelType == "so-vits-svc-40v2":
             return self.voiceChanger.loadModel(config, pyTorch_model_file, onnx_model_file, clusterTorchModel)
         elif self.modelType == "RVC":
-            return self.voiceChanger.loadModel(config, pyTorch_model_file, onnx_model_file, feature_file, index_file)
+            return self.voiceChanger.loadModel(config, pyTorch_model_file, onnx_model_file, feature_file, index_file, is_half)
         else:
             return self.voiceChanger.loadModel(config, pyTorch_model_file, onnx_model_file, clusterTorchModel)
 

server/voice_changer/VoiceChangerManager.py

@@ -10,8 +10,8 @@ class VoiceChangerManager():
             cls._instance.voiceChanger = VoiceChanger(params)
         return cls._instance
 
-    def loadModel(self, config, model, onnx_model, clusterTorchModel, feature_file, index_file):
+    def loadModel(self, config, model, onnx_model, clusterTorchModel, feature_file, index_file, is_half: bool = True):
-        info = self.voiceChanger.loadModel(config, model, onnx_model, clusterTorchModel, feature_file, index_file)
+        info = self.voiceChanger.loadModel(config, model, onnx_model, clusterTorchModel, feature_file, index_file, is_half)
         info["status"] = "OK"
         return info