sample update

client/demo/src/components/demo/components/201_StartButtonRow.tsx

@@ -70,7 +70,7 @@ export const StartButtonRow = (_props: StartButtonRowProps) => {
                 </div>
             </div>
         )
-    }, [guiState.isConverting, appState.clientSetting.start, appState.clientSetting.stop, appState.serverSetting.serverSetting, , appState.serverSetting.updateServerSettings])
+    }, [guiState.isConverting, appState.clientSetting.start, appState.clientSetting.stop, appState.serverSetting.serverSetting, appState.serverSetting.updateServerSettings])
 
     return startButtonRow
 }

client/demo/src/components/demo/components/204_ModelSwitchRow.tsx

@@ -39,7 +39,7 @@ export const ModelSwitchRow = (_props: ModelSwitchRowProps) => {
                 if (x.modelType == "pyTorchRVC" || x.modelType == "pyTorchRVCNono") {
                     return "org"
                 } else if (x.modelType == "pyTorchRVCv2" || x.modelType == "pyTorchRVCv2Nono") {
-                    return "orv_v2"
+                    return "org_v2"
                 } else if (x.modelType == "pyTorchWebUI" || x.modelType == "pyTorchWebUINono") {
                     return "webui"
                 } else {

client/demo/src/components/demo/components/207_ServerOpertationRow.tsx

@@ -15,6 +15,7 @@ export const ServerOpertationRow = (props: ServerOpertationRowProps) => {
     const guiState = useGuiState()
 
     const serverOpertationRow = useMemo(() => {
+        console.log("UPDATE OPERATION ROW:::::::::::::::::::::::")
 
         const onnxExportButtonAction = async () => {
 
@@ -78,7 +79,7 @@ export const ServerOpertationRow = (props: ServerOpertationRowProps) => {
                 </div>
             </>
         )
-    }, [appState.getInfo, appState.serverSetting.serverSetting])
+    }, [appState.getInfo, appState.serverSetting.serverSetting, guiState.isConverting])
 
     return serverOpertationRow
 }

client/demo/src/components/demo/components/a01_MergeLab.Row.tsx

@@ -121,7 +121,7 @@ export const MergeLabRow = (_props: MergeLabRowProps) => {
                 if (modelInfo.modelType == "pyTorchRVC" || modelInfo.modelType == "pyTorchRVCNono") {
                     return "org"
                 } else if (modelInfo.modelType == "pyTorchRVCv2" || modelInfo.modelType == "pyTorchRVCv2Nono") {
-                    return "orv_v2"
+                    return "g_v2"
                 } else if (modelInfo.modelType == "pyTorchWebUI" || modelInfo.modelType == "pyTorchWebUINono") {
                     return "webui"
                 } else {

server/ModelSample.py

@@ -12,7 +12,6 @@ class RVCModelSample:
     name: str = ""
     modelUrl: str = ""
     indexUrl: str = ""
-    featureUrl: str = ""
     termsOfUseUrl: str = ""
     credit: str = ""
     description: str = ""

server/const.py

@@ -101,8 +101,8 @@ class ServerAudioDeviceTypes(Enum):
 
 
 SAMPLES_JSONS = [
-    "https://huggingface.co/wok000/vcclient_model/raw/main/samples_0001.json",
+    "https://huggingface.co/wok000/vcclient_model/raw/main/samples_0003_t.json",
-    "https://huggingface.co/wok000/vcclient_model/raw/main/samples_0002.json",
+    "https://huggingface.co/wok000/vcclient_model/raw/main/samples_0003_o.json",
 ]
 RVC_MODEL_DIRNAME = "rvc"
 RVC_MAX_SLOT_NUM = 5

server/samples_0003_o.json

@@ -0,0 +1,74 @@
+{
+  "RVC": [
+    {
+      "id": "KikotoKurage_o",
+      "lang": "ja-JP",
+      "tag": ["v2", "onnx"],
+      "name": "黄琴海月",
+      "modelUrl": "https://huggingface.co/wok000/vcclient_model/resolve/main/rvc_v2_alpha/kikoto_kurage/kikoto_kurage_v2_40k_e100_simple.onnx",
+      "indexUrl": "https://huggingface.co/wok000/vcclient_model/resolve/main/rvc_v2_alpha/kikoto_kurage/added_IVF5181_Flat_nprobe_1_v2.index.bin",
+      "termsOfUseUrl": "https://huggingface.co/wok000/vcclient_model/raw/main/rvc_v2_alpha/kikoto_kurage/terms_of_use.txt",
+      "credit": "黄琴海月",
+      "description": "",
+      "sampleRate": 40000,
+      "modelType": "rvc_v2",
+      "f0": true
+    },
+    {
+      "id": "KikotoMahiro_o",
+      "lang": "ja-JP",
+      "tag": ["v2", "onnx"],
+      "name": "黄琴まひろ",
+      "modelUrl": "https://huggingface.co/wok000/vcclient_model/resolve/main/rvc_v2_alpha/kikoto_mahiro/kikoto_mahiro_v2_40k_simple.onnx",
+      "indexUrl": "https://huggingface.co/wok000/vcclient_model/resolve/main/rvc_v2_alpha/kikoto_mahiro/added_IVF6881_Flat_nprobe_1_v2.index.bin",
+      "termsOfUseUrl": "",
+      "credit": "黄琴まひろ",
+      "description": "",
+      "sampleRate": 40000,
+      "modelType": "rvc_v2",
+      "f0": true
+    },
+    {
+      "id": "TokinaShigure_o",
+      "lang": "ja-JP",
+      "tag": ["v2", "onnx"],
+      "name": "刻鳴時雨",
+      "modelUrl": "https://huggingface.co/wok000/vcclient_model/resolve/main/rvc_v2_alpha/tokina_shigure/tokina_shigure_v2_40k_e100_simple.onnx",
+      "indexUrl": "https://huggingface.co/wok000/vcclient_model/resolve/main/rvc_v2_alpha/tokina_shigure/added_IVF2736_Flat_nprobe_1_v2.index.bin",
+      "termsOfUseUrl": "https://huggingface.co/wok000/vcclient_model/raw/main/rvc_v2_alpha/tokina_shigure/terms_of_use.txt",
+      "credit": "刻鳴時雨",
+      "description": "",
+      "sampleRate": 40000,
+      "modelType": "rvc_v2",
+      "f0": true
+    },
+    {
+      "id": "Amitaro_o",
+      "lang": "ja-JP",
+      "tag": ["v2", "onnx"],
+      "name": "あみたろ",
+      "modelUrl": "https://huggingface.co/wok000/vcclient_model/resolve/main/rvc_v2_alpha/amitaro/amitaro_v2_40k_e100_simple.onnx",
+      "indexUrl": "https://huggingface.co/wok000/vcclient_model/resolve/main/rvc_v2_alpha/amitaro/added_IVF3139_Flat_nprobe_1_v2.index.bin",
+      "termsOfUseUrl": "https://huggingface.co/wok000/vcclient_model/raw/main/rvc_v2_alpha/amitaro/terms_of_use.txt",
+      "credit": "あみたろ",
+      "description": "",
+      "sampleRate": 40000,
+      "modelType": "rvc_v2",
+      "f0": true
+    },
+    {
+      "id": "Tsukuyomi-chan_o",
+      "lang": "ja-JP",
+      "tag": ["v2", "onnx"],
+      "name": "つくよみちゃん",
+      "modelUrl": "https://huggingface.co/wok000/vcclient_model/resolve/main/rvc_v2_alpha/tsukuyomi-chan/tsukuyomi_v2_40k_e100_simple.onnx",
+      "indexUrl": "https://huggingface.co/wok000/vcclient_model/resolve/main/rvc_v2_alpha/tsukuyomi-chan/added_IVF7852_Flat_nprobe_1_v2.index.bin",
+      "termsOfUseUrl": "https://huggingface.co/wok000/vcclient_model/blob/main/rvc_v2_alpha/tsukuyomi-chan/terms_of_use.txt",
+      "credit": "つくよみちゃん",
+      "description": "",
+      "sampleRate": 40000,
+      "modelType": "rvc_v2",
+      "f0": true
+    }
+  ]
+}

server/samples_0003_t.json

@@ -0,0 +1,74 @@
+{
+  "RVC": [
+    {
+      "id": "KikotoKurage_t",
+      "lang": "ja-JP",
+      "tag": ["v2", "torch"],
+      "name": "黄琴海月",
+      "modelUrl": "https://huggingface.co/wok000/vcclient_model/resolve/main/rvc_v2_alpha/kikoto_kurage/kikoto_kurage_v2_40k_e100.pth",
+      "indexUrl": "https://huggingface.co/wok000/vcclient_model/resolve/main/rvc_v2_alpha/kikoto_kurage/added_IVF5181_Flat_nprobe_1_v2.index.bin",
+      "termsOfUseUrl": "https://huggingface.co/wok000/vcclient_model/raw/main/rvc_v2_alpha/kikoto_kurage/terms_of_use.txt",
+      "credit": "黄琴海月",
+      "description": "",
+      "sampleRate": 40000,
+      "modelType": "rvc_v2",
+      "f0": true
+    },
+    {
+      "id": "KikotoMahiro_t",
+      "lang": "ja-JP",
+      "tag": ["v2", "torch"],
+      "name": "黄琴まひろ",
+      "modelUrl": "https://huggingface.co/wok000/vcclient_model/resolve/main/rvc_v2_alpha/kikoto_mahiro/kikoto_mahiro_v2_40k.pth",
+      "indexUrl": "https://huggingface.co/wok000/vcclient_model/resolve/main/rvc_v2_alpha/kikoto_mahiro/added_IVF6881_Flat_nprobe_1_v2.index.bin",
+      "termsOfUseUrl": "",
+      "credit": "黄琴まひろ",
+      "description": "",
+      "sampleRate": 40000,
+      "modelType": "rvc_v2",
+      "f0": true
+    },
+    {
+      "id": "TokinaShigure_t",
+      "lang": "ja-JP",
+      "tag": ["v2", "torch"],
+      "name": "刻鳴時雨",
+      "modelUrl": "https://huggingface.co/wok000/vcclient_model/resolve/main/rvc_v2_alpha/tokina_shigure/tokina_shigure_v2_40k_e100.pth",
+      "indexUrl": "https://huggingface.co/wok000/vcclient_model/resolve/main/rvc_v2_alpha/tokina_shigure/added_IVF2736_Flat_nprobe_1_v2.index.bin",
+      "termsOfUseUrl": "https://huggingface.co/wok000/vcclient_model/raw/main/rvc_v2_alpha/tokina_shigure/terms_of_use.txt",
+      "credit": "刻鳴時雨",
+      "description": "",
+      "sampleRate": 40000,
+      "modelType": "rvc_v2",
+      "f0": true
+    },
+    {
+      "id": "Amitaro_t",
+      "lang": "ja-JP",
+      "tag": ["v2", "torch"],
+      "name": "あみたろ",
+      "modelUrl": "https://huggingface.co/wok000/vcclient_model/resolve/main/rvc_v2_alpha/amitaro/amitaro_v2_40k_e100.pth",
+      "indexUrl": "https://huggingface.co/wok000/vcclient_model/resolve/main/rvc_v2_alpha/amitaro/added_IVF3139_Flat_nprobe_1_v2.index.bin",
+      "termsOfUseUrl": "https://huggingface.co/wok000/vcclient_model/raw/main/rvc_v2_alpha/amitaro/terms_of_use.txt",
+      "credit": "あみたろ",
+      "description": "",
+      "sampleRate": 40000,
+      "modelType": "rvc_v2",
+      "f0": true
+    },
+    {
+      "id": "Tsukuyomi-chan_t",
+      "lang": "ja-JP",
+      "tag": ["v2", "torch"],
+      "name": "つくよみちゃん",
+      "modelUrl": "https://huggingface.co/wok000/vcclient_model/resolve/main/rvc_v2_alpha/tsukuyomi-chan/tsukuyomi_v2_40k_e100.pth",
+      "indexUrl": "https://huggingface.co/wok000/vcclient_model/resolve/main/rvc_v2_alpha/tsukuyomi-chan/added_IVF7852_Flat_nprobe_1_v2.index.bin",
+      "termsOfUseUrl": "https://huggingface.co/wok000/vcclient_model/blob/main/rvc_v2_alpha/tsukuyomi-chan/terms_of_use.txt",
+      "credit": "つくよみちゃん",
+      "description": "",
+      "sampleRate": 40000,
+      "modelType": "rvc_v2",
+      "f0": true
+    }
+  ]
+}

server/voice_changer/RVC/ModelSlot.py

@@ -17,7 +17,7 @@ class ModelSlot:
     samplingRate: int = -1
     f0: bool = True
     embChannels: int = 256
-    embOutputLayter: int = 9
+    embOutputLayer: int = 9
     useFinalProj: bool = True
     deprecated: bool = False
     embedder: EnumEmbedderTypes = EnumEmbedderTypes.hubert

server/voice_changer/RVC/ModelSlotGenerator.py

@@ -66,7 +66,7 @@ def _setInfoByPytorch(slot: ModelSlot):
                 else EnumInferenceTypes.pyTorchRVCNono
             )
             slot.embChannels = 256
-            slot.embOutputLayter = 9
+            slot.embOutputLayer = 9
             slot.useFinalProj = True
             slot.embedder = EnumEmbedderTypes.hubert
         else:
@@ -76,7 +76,7 @@ def _setInfoByPytorch(slot: ModelSlot):
                 else EnumInferenceTypes.pyTorchRVCv2Nono
             )
             slot.embChannels = 768
-            slot.embOutputLayter = 12
+            slot.embOutputLayer = 12
             slot.useFinalProj = False
             slot.embedder = EnumEmbedderTypes.hubert
 
@@ -89,7 +89,7 @@ def _setInfoByPytorch(slot: ModelSlot):
             else EnumInferenceTypes.pyTorchWebUINono
         )
         slot.embChannels = cpt["config"][17]
-        slot.embOutputLayter = (
+        slot.embOutputLayer = (
             cpt["embedder_output_layer"] if "embedder_output_layer" in cpt else 9
         )
         if slot.embChannels == 256:
@@ -100,19 +100,19 @@ def _setInfoByPytorch(slot: ModelSlot):
         # DDPNモデルの情報を表示
         if (
             slot.embChannels == 256
-            and slot.embOutputLayter == 9
+            and slot.embOutputLayer == 9
             and slot.useFinalProj is True
         ):
             print("[Voice Changer] DDPN Model: Original v1 like")
         elif (
             slot.embChannels == 768
-            and slot.embOutputLayter == 12
+            and slot.embOutputLayer == 12
             and slot.useFinalProj is False
         ):
             print("[Voice Changer] DDPN Model: Original v2 like")
         else:
             print(
-                f"[Voice Changer] DDPN Model: ch:{slot.embChannels}, L:{slot.embOutputLayter}, FP:{slot.useFinalProj}"
+                f"[Voice Changer] DDPN Model: ch:{slot.embChannels}, L:{slot.embOutputLayer}, FP:{slot.useFinalProj}"
             )
 
         slot.embedder = cpt["embedder_name"]
@@ -144,10 +144,15 @@ def _setInfoByONNX(slot: ModelSlot):
         # slot.modelType = metadata["modelType"]
         slot.embChannels = metadata["embChannels"]
 
-        slot.embOutputLayter = (
+        slot.embOutputLayer = (
-            metadata["embedder_output_layer"]
+            metadata["embOutputLayer"] if "embOutputLayer" in metadata else 9
-            if "embedder_output_layer" in metadata
+        )
-            else 9
+        slot.useFinalProj = (
+            metadata["useFinalProj"]
+            if "useFinalProj" in metadata
+            else True
+            if slot.embChannels == 256
+            else False
         )
 
         if slot.embChannels == 256:
@@ -155,7 +160,23 @@ def _setInfoByONNX(slot: ModelSlot):
         else:
             slot.useFinalProj = False
 
-        print("ONNX", slot)
+        # ONNXモデルの情報を表示
+        if (
+            slot.embChannels == 256
+            and slot.embOutputLayer == 9
+            and slot.useFinalProj is True
+        ):
+            print("[Voice Changer] ONNX Model: Original v1 like")
+        elif (
+            slot.embChannels == 768
+            and slot.embOutputLayer == 12
+            and slot.useFinalProj is False
+        ):
+            print("[Voice Changer] ONNX Model: Original v2 like")
+        else:
+            print(
+                f"[Voice Changer] ONNX Model: ch:{slot.embChannels}, L:{slot.embOutputLayer}, FP:{slot.useFinalProj}"
+            )
 
         if "embedder" not in metadata:
             slot.embedder = EnumEmbedderTypes.hubert

server/voice_changer/RVC/RVC.py

@@ -86,7 +86,7 @@ class RVC:
                 if len(slot.modelFile) > 0:
                     self.prepareModel(i)
                     self.settings.modelSlotIndex = i
-                    self.switchModel()
+                    self.switchModel(self.settings.modelSlotIndex)
                     self.initialLoad = False
                     break
 
@@ -115,12 +115,10 @@ class RVC:
             if sampleInfo is None:
                 print("[Voice Changer] sampleInfo is None")
                 return
-            modelPath, indexPath, featurePath = downloadModelFiles(sampleInfo)
+            modelPath, indexPath = downloadModelFiles(sampleInfo)
             params["files"]["rvcModel"] = modelPath
             if indexPath is not None:
                 params["files"]["rvcIndex"] = indexPath
-            if featurePath is not None:
-                params["files"]["rvcFeature"] = featurePath
             params["credit"] = sampleInfo.credit
             params["description"] = sampleInfo.description
             params["name"] = sampleInfo.name
@@ -152,7 +150,7 @@ class RVC:
         if self.initialLoad:
             self.prepareModel(target_slot_idx)
             self.settings.modelSlotIndex = target_slot_idx
-            self.switchModel()
+            self.switchModel(self.settings.modelSlotIndex)
             self.initialLoad = False
         elif target_slot_idx == self.currentSlot:
             self.prepareModel(target_slot_idx)
@@ -242,11 +240,12 @@ class RVC:
         self.next_index_ratio = modelSlot.defaultIndexRatio
         self.next_samplingRate = modelSlot.samplingRate
         self.next_framework = "ONNX" if modelSlot.isONNX else "PyTorch"
-        self.needSwitch = True
+        # self.needSwitch = True
         print("[Voice Changer] Prepare done.")
+        self.switchModel(slot)
         return self.get_info()
 
-    def switchModel(self):
+    def switchModel(self, slot: int):
         print("[Voice Changer] Switching model..")
         self.pipeline = self.next_pipeline
         self.settings.tran = self.next_trans
@@ -254,6 +253,8 @@ class RVC:
         self.settings.modelSamplingRate = self.next_samplingRate
         self.settings.framework = self.next_framework
 
+        # self.currentSlot = self.settings.modelSlotIndex # prepareModelから呼ばれるということはupdate_settingsの中で呼ばれるということなので、まだmodelSlotIndexは更新されていない
+        self.currentSlot = slot
         print(
             "[Voice Changer] Switching model..done",
         )
@@ -308,13 +309,12 @@ class RVC:
                 self.currentSlot,
             )
             raise NoModeLoadedException("model_common")
-        if self.needSwitch:
+        # if self.needSwitch:
-            print(
+        #     print(
-                f"[Voice Changer] Switch model {self.currentSlot} -> {self.settings.modelSlotIndex}"
+        #         f"[Voice Changer] Switch model {self.currentSlot} -> {self.settings.modelSlotIndex}"
-            )
+        #     )
-            self.currentSlot = self.settings.modelSlotIndex
+        #     self.switchModel()
-            self.switchModel()
+        #     self.needSwitch = False
-            self.needSwitch = False
 
         half = self.deviceManager.halfPrecisionAvailable(self.settings.gpu)
 
@@ -333,7 +333,7 @@ class RVC:
         f0_up_key = self.settings.tran
         index_rate = self.settings.indexRatio
         if_f0 = 1 if self.settings.modelSlots[self.currentSlot].f0 else 0
-        embOutputLayer = self.settings.modelSlots[self.currentSlot].embOutputLayter
+        embOutputLayer = self.settings.modelSlots[self.currentSlot].embOutputLayer
         useFinalProj = self.settings.modelSlots[self.currentSlot].useFinalProj
 
         audio_out = self.pipeline.exec(

server/voice_changer/RVC/SampleDownloader.py

@@ -16,10 +16,10 @@ def checkRvcModelExist(model_dir: str):
 
 def downloadInitialSampleModels(sampleJsons: list[str], model_dir: str):
     sampleModelIds = [
-        "KikotoMahiro",
+        "KikotoMahiro_o",
-        "TokinaShigure",
+        "TokinaShigure_o",
-        "Amitaro",
+        "Amitaro_o",
-        "Tsukuyomi-chan",
+        "Tsukuyomi-chan_o",
     ]
     sampleModels = getModelSamples(sampleJsons, "RVC")
     if sampleModels is None:
@@ -60,23 +60,7 @@ def downloadInitialSampleModels(sampleJsons: list[str], model_dir: str):
                         "position": line_num,
                     }
                 )
-                sampleParams["files"]["rvcIndex"] = modelFilePath
+                sampleParams["files"]["rvcIndex"] = indexPath
-                line_num += 1
-
-            featurePath = None
-            if hasattr(sample, "featureUrl") or sample.featureUrl != "":
-                featurePath = os.path.join(
-                    slotDir,
-                    os.path.basename(sample.featureUrl),
-                )
-                downloadParams.append(
-                    {
-                        "url": sample.featureUrl,
-                        "saveTo": featurePath,
-                        "position": line_num,
-                    }
-                )
-                sampleParams["files"]["rvcFeatur"] = modelFilePath
                 line_num += 1
 
             sampleParams["sampleId"] = sample.id
@@ -123,19 +107,8 @@ def downloadModelFiles(sampleInfo: RVCModelSample):
             }
         )
 
-    featurePath = None
-    if hasattr(sampleInfo, "featureUrl") or sampleInfo.featureUrl != "":
-        featurePath = os.path.join(TMP_DIR, os.path.basename(sampleInfo.featureUrl))
-        downloadParams.append(
-            {
-                "url": sampleInfo.featureUrl,
-                "saveTo": featurePath,
-                "position": 2,
-            }
-        )
-
     print("[Voice Changer] Downloading model files...", end="")
     with ThreadPoolExecutor() as pool:
         pool.map(download_no_tqdm, downloadParams)
     print("")
-    return modelPath, indexPath, featurePath
+    return modelPath, indexPath

server/voice_changer/RVC/modelMerger/MergeModel.py

@@ -64,5 +64,10 @@ def merge_model(request: MergeModelRequest):
     merged["embedder_name"] = (
         state_dict["embedder_name"] if "embedder_name" in state_dict else None
     )
+    merged["embedder_output_layer"] = (
+        state_dict["embedder_output_layer"]
+        if "embedder_output_layer" in state_dict
+        else None
+    )
     print("write metadata done.")
     return merged

server/voice_changer/RVC/onnxExporter/export2onnx.py

@@ -45,6 +45,8 @@ def export2onnx(gpu: int, modelSlot: ModelSlot):
         "embChannels": modelSlot.embChannels,
         # ↓EnumEmbedderTypesのままだとシリアライズできないのでテキスト化
         "embedder": modelSlot.embedder.value,
+        "embOutputLayer": modelSlot.embOutputLayer,
+        "useFinalProj": modelSlot.useFinalProj,
     }
     gpuMomory = DeviceManager.get_instance().getDeviceMemory(gpu)
     print(f"[Voice Changer] exporting onnx... gpu_id:{gpu} gpu_mem:{gpuMomory}")

server/voice_changer/RVC/pipeline/PipelineGenerator.py

@@ -74,6 +74,7 @@ def _loadIndex(modelSlot: ModelSlot):
         return None
 
     try:
+        print("Try loading...", modelSlot.indexFile)
         index = faiss.read_index(modelSlot.indexFile)
     except:
         print("[Voice Changer] load index failed. Use no index.")