PublicMirrors/voice-changer
1
0
Fork 0
mirror of https://github.com/w-okada/voice-changer.git synced 2025-02-02 16:23:58 +03:00
Code Issues Actions Packages Projects Releases Wiki Activity

fix model upload bug

Browse Source
This commit is contained in:
wataru 2023-01-08 21:56:39 +09:00
parent 0109f2671f
commit ef89e9aefa
5 changed files with 30 additions and 17 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
client/demo/src/100_options_microphone.tsx
View File

@ -142,10 +142,23 @@ export const useMicrophoneOptions = () => {
await clientState.loadModel(serverSetting.configFile, serverSetting.pyTorchModel, serverSetting.onnxModel)
setUploadProgress(0)
setIsUploading(false)
// await clientState.updateSettings(ServerSettingKey.framework, serverSetting.framework)
// await clientState.updateSettings(ServerSettingKey.onnxExecutionProvider, serverSetting.onnxExecutionProvider)
// await clientState.updateSettings(ServerSettingKey.srcId, speakerSetting.srcId)
// await clientState.updateSettings(ServerSettingKey.dstId, speakerSetting.dstId)
// await clientState.updateSettings(ServerSettingKey.gpu, convertSetting.gpu)
// await clientState.updateSettings(ServerSettingKey.crossFadeOffsetRate, convertSetting.crossFadeOffsetRate)
// await clientState.updateSettings(ServerSettingKey.crossFadeEndRate, convertSetting.crossFadeEndRate)
}
}
setLoadModelFunc(loadModel)
}, [serverSetting.configFile, serverSetting.pyTorchModel, serverSetting.onnxModel])
}, [serverSetting.configFile, serverSetting.pyTorchModel, serverSetting.onnxModel,
serverSetting.framework, serverSetting.onnxExecutionProvider, speakerSetting.srcId, speakerSetting.dstId, convertSetting.gpu, convertSetting.crossFadeOffsetRate, convertSetting.crossFadeEndRate
])
const voiceChangerSetting = useMemo(() => {

6
client/demo/src/hooks/useClient.ts
View File

@ -155,17 +155,19 @@ export const useClient = (props: UseClientProps): ClientState => {
const uploadFile = useMemo(() => {
return async (file: File, onprogress: (progress: number, end: boolean) => void) => {
await initializedPromise
console.log("uploaded.....")
const num = await voiceChangerClientRef.current.uploadFile(file, onprogress)
const res = await voiceChangerClientRef.current.concatUploadedFile(file, num)
console.log("upload", num, res)
console.log("uploaded", num, res)
}
}, [])
const loadModel = useMemo(() => {
return async (configFile: File, pyTorchModelFile: File | null, onnxModelFile: File | null) => {
console.log("load model....")
await initializedPromise
await voiceChangerClientRef.current.loadModel(configFile, pyTorchModelFile, onnxModelFile)
console.log("load model")
console.log("loaded model")
}
}, [])

2
client/lib/src/ServerConfigurator.ts
View File

@ -96,7 +96,7 @@ export class ServerConfigurator {
concatUploadedFile = async (file: File, chunkNum: number) => {
const url = this.serverUrl + "/concat_uploaded_file"
new Promise<void>((resolve) => {
await new Promise<void>((resolve) => {
const formData = new FormData();
formData.append("filename", file.name);
formData.append("filenameChunkNum", "" + chunkNum);

18
server/voice_changer/VoiceChanger.py
View File

@ -36,9 +36,9 @@ class VocieChangerSettings():
class VoiceChanger():
def __init__(self, config:str, pyTorch_model_file:str=None, onnx_model_file:str=None):
def __init__(self, config:str):
# 初期化
self.settings = VocieChangerSettings(config_file=config, pyTorch_model_file=pyTorch_model_file, onnx_model_file=onnx_model_file)
self.settings = VocieChangerSettings(config_file=config)
self.unpackedData_length=0
# 共通で使用する情報を収集
self.hps = utils.get_hparams_from_file(config)
@ -53,6 +53,11 @@ 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):
self.settings.config_file = config
self.settings.pyTorch_model_file = pyTorch_model_file
self.settings.onnx_model_file = onnx_model_file
# PyTorchモデル生成
if pyTorch_model_file != None:
self.net_g = SynthesizerTrn(
@ -70,21 +75,14 @@ class VoiceChanger():
if onnx_model_file != None:
ort_options = onnxruntime.SessionOptions()
ort_options.intra_op_num_threads = 8
# ort_options.execution_mode = onnxruntime.ExecutionMode.ORT_SEQUENTIAL
# ort_options.execution_mode = onnxruntime.ExecutionMode.ORT_PARALLEL
# ort_options.inter_op_num_threads = 8
self.onnx_session = onnxruntime.InferenceSession(
onnx_model_file,
providers=providers
)
# print("ONNX_MDEOL!1", self.onnx_session.get_providers())
# self.onnx_session.set_providers(providers=["CPUExecutionProvider"])
# print("ONNX_MDEOL!1", self.onnx_session.get_providers())
# self.onnx_session.set_providers(providers=["DmlExecutionProvider"])
# print("ONNX_MDEOL!1", self.onnx_session.get_providers())
else:
self.onnx_session = None
def destroy(self):
if hasattr(self, "net_g"):
del self.net_g

6
server/voice_changer/VoiceChangerManager.py
View File

@ -11,9 +11,9 @@ class VoiceChangerManager():
return cls._instance
def loadModel(self, config, model, onnx_model):
if hasattr(self, 'voiceChanger') == True:
self.voiceChanger.destroy()
self.voiceChanger = VoiceChanger(config, model, onnx_model)
if hasattr(self, 'voiceChanger') == False:
self.voiceChanger = VoiceChanger(config)
self.voiceChanger.loadModel(config, model, onnx_model)
def get_info(self):
if hasattr(self, 'voiceChanger'):