update

demo/MMVCServerSIO.py

@@ -0,0 +1,280 @@
+import sys, os, struct, argparse, logging, shutil,  base64, traceback
+sys.path.append("/MMVC_Trainer")
+sys.path.append("/MMVC_Trainer/text")
+
+import uvicorn
+from fastapi import FastAPI, UploadFile, File, Form
+from fastapi.middleware.cors import CORSMiddleware
+from fastapi.responses import JSONResponse
+from fastapi.encoders import jsonable_encoder
+from pydantic import BaseModel
+
+from scipy.io.wavfile import write, read
+
+import socketio
+from distutils.util import strtobool
+from datetime import datetime
+
+import torch
+import numpy as np
+
+from mods.ssl import create_self_signed_cert
+from mods.VoiceChanger import VoiceChanger
+
+class UvicornSuppressFilter(logging.Filter):
+    def filter(self, record):
+        return False
+
+logger = logging.getLogger("uvicorn.error")
+logger.addFilter(UvicornSuppressFilter())
+# logger.propagate = False
+logger = logging.getLogger("multipart.multipart")
+logger.propagate = False
+
+
+
+class VoiceModel(BaseModel):
+    gpu: int
+    srcId: int
+    dstId: int
+    timestamp: int
+    buffer: str
+
+
+class MyCustomNamespace(socketio.AsyncNamespace): 
+    def __init__(self, namespace):
+        super().__init__(namespace)
+
+    def loadModel(self, config, model):
+        if hasattr(self, 'voiceChanger') == True:
+            self.voiceChanger.destroy()
+        self.voiceChanger = VoiceChanger(config, model)
+
+    def changeVoice(self, gpu, srcId, dstId, timestamp, unpackedData):
+        return self.voiceChanger.on_request(gpu, srcId, dstId, timestamp, unpackedData)
+
+    def on_connect(self, sid, environ):
+        # print('[{}] connet sid : {}'.format(datetime.now().strftime('%Y-%m-%d %H:%M:%S') , sid))
+        pass
+
+    async def on_request_message(self, sid, msg): 
+        # print("on_request_message", torch.cuda.memory_allocated())
+        gpu = int(msg[0])
+        srcId = int(msg[1])
+        dstId = int(msg[2])
+        timestamp = int(msg[3])
+        data = msg[4]
+        # print(srcId, dstId, timestamp)
+        unpackedData = np.array(struct.unpack('<%sh'%(len(data) // struct.calcsize('<h') ), data))
+        audio1 = self.changeVoice(gpu, srcId, dstId, timestamp, unpackedData)
+
+        bin = struct.pack('<%sh'%len(audio1), *audio1)
+
+        await self.emit('response',[timestamp, bin])
+
+    def on_disconnect(self, sid):
+        # print('[{}] disconnect'.format(datetime.now().strftime('%Y-%m-%d %H:%M:%S')))
+        pass;
+
+
+def setupArgParser():
+    parser = argparse.ArgumentParser()
+    parser.add_argument("-p", type=int, default=8080, help="port")
+    parser.add_argument("-c", type=str, help="path for the config.json")
+    parser.add_argument("-m", type=str, help="path for the model file")
+    parser.add_argument("--https", type=strtobool, default=False, help="use https")
+    parser.add_argument("--httpsKey", type=str, default="ssl.key", help="path for the key of https")
+    parser.add_argument("--httpsCert", type=str, default="ssl.cert", help="path for the cert of https")
+    parser.add_argument("--httpsSelfSigned", type=strtobool, default=True, help="generate self-signed certificate")
+    return parser
+
+def printMessage(message, level=0):
+    if level == 0:
+        print(f"\033[17m{message}\033[0m")
+    elif level == 1:
+        print(f"\033[34m    {message}\033[0m")
+    elif level == 2:
+        print(f"\033[32m    {message}\033[0m")
+    else:
+        print(f"\033[47m    {message}\033[0m")
+
+global app_socketio
+
+
+printMessage(f"Phase name:{__name__}", level=2)
+thisFilename = os.path.basename(__file__)[:-3]
+
+
+if __name__ == thisFilename:
+    printMessage(f"PHASE3:{__name__}", level=2)
+    parser = setupArgParser()
+    args = parser.parse_args()
+    PORT = args.p
+    CONFIG = args.c
+    MODEL  = args.m
+
+    app_fastapi = FastAPI()
+    sio = socketio.AsyncServer(
+        async_mode='asgi',
+        cors_allowed_origins='*'
+    )
+    namespace = MyCustomNamespace('/test')
+    sio.register_namespace(namespace) 
+    if CONFIG and MODEL:
+        namespace.loadModel(CONFIG, MODEL)
+    app_socketio = socketio.ASGIApp(
+        sio, 
+        other_asgi_app=app_fastapi,
+        static_files={
+            '': '../frontend/dist',
+            '/': '../frontend/dist/index.html',
+        }
+    )
+
+    @app_fastapi.get("/api/hello")
+    async def index():
+        return {"result": "Index"}
+    
+
+    @app_fastapi.post("/api/uploadfile/model")
+    async def upload_file(configFile:UploadFile = File(...), modelFile: UploadFile = File(...)):
+        if configFile and modelFile:
+            for file in [modelFile, configFile]:
+                filename = file.filename
+                fileobj = file.file
+                upload_dir = open(os.path.join(".", filename),'wb+')
+                shutil.copyfileobj(fileobj, upload_dir)
+                upload_dir.close()
+            namespace.loadModel(configFile.filename, modelFile.filename)                
+            return {"uploaded files": f"{configFile.filename}, {modelFile.filename} "}
+        return {"Error": "uploaded file is not found."}
+
+
+
+    @app_fastapi.post("/test")
+    async def post_test(voice:VoiceModel):
+        try:
+            # print("POST REQUEST PROCESSING....")
+            gpu = voice.gpu
+            srcId = voice.srcId
+            dstId = voice.dstId
+            timestamp = voice.timestamp
+            buffer = voice.buffer
+            wav = base64.b64decode(buffer)
+
+            if wav==0:
+                samplerate, data=read("dummy.wav")
+                unpackedData = data
+            else:
+                unpackedData = np.array(struct.unpack('<%sh'%(len(wav) // struct.calcsize('<h') ), wav))
+                write("logs/received_data.wav", 24000, unpackedData.astype(np.int16))
+
+            changedVoice = namespace.changeVoice(gpu, srcId, dstId, timestamp, unpackedData)
+            changedVoiceBase64 = base64.b64encode(changedVoice).decode('utf-8')
+
+            data = {
+                "gpu":gpu,
+                "srcId":srcId,
+                "dstId":dstId,
+                "timestamp":timestamp,
+                "changedVoiceBase64":changedVoiceBase64
+            }
+
+            json_compatible_item_data = jsonable_encoder(data)
+            
+            return JSONResponse(content=json_compatible_item_data)
+        except Exception as e:
+            print("REQUEST PROCESSING!!!! EXCEPTION!!!", e)
+            print(traceback.format_exc())
+            return str(e)
+
+
+if __name__ == '__mp_main__':
+    printMessage(f"PHASE2:{__name__}", level=2)
+
+
+if __name__ == '__main__':
+    printMessage(f"PHASE1:{__name__}", level=2)
+
+    parser = setupArgParser()
+    args = parser.parse_args()
+    PORT = args.p
+    CONFIG = args.c
+    MODEL  = args.m
+
+    printMessage(f"Start MMVC SocketIO Server", level=0)
+    printMessage(f"CONFIG:{CONFIG}, MODEL:{MODEL}", level=1)
+
+    if os.environ["EX_PORT"]:
+        EX_PORT = os.environ["EX_PORT"]
+        printMessage(f"External_Port:{EX_PORT} Internal_Port:{PORT}", level=1)
+    else:
+        printMessage(f"Internal_Port:{PORT}", level=1)
+
+    if os.environ["EX_IP"]:
+        EX_IP = os.environ["EX_IP"]
+        printMessage(f"External_IP:{EX_IP}", level=1)
+
+    # HTTPS key/cert作成
+    if args.https and args.httpsSelfSigned == 1:
+        # HTTPS(おれおれ証明書生成) 
+        os.makedirs("./key", exist_ok=True)
+        key_base_name = f"{datetime.now().strftime('%Y%m%d_%H%M%S')}"
+        keyname = f"{key_base_name}.key"
+        certname = f"{key_base_name}.cert"
+        create_self_signed_cert(certname, keyname, certargs=
+                            {"Country": "JP",
+                                "State": "Tokyo",
+                                "City": "Chuo-ku",
+                                "Organization": "F",
+                                "Org. Unit": "F"}, cert_dir="./key")
+        key_path = os.path.join("./key", keyname)
+        cert_path = os.path.join("./key", certname)
+        printMessage(f"protocol: HTTPS(self-signed), key:{key_path}, cert:{cert_path}", level=1)
+    elif args.https and args.httpsSelfSigned == 0:
+        # HTTPS 
+        key_path = args.httpsKey
+        cert_path = args.httpsCert
+        printMessage(f"protocol: HTTPS, key:{key_path}, cert:{cert_path}", level=1)
+    else:
+        # HTTP
+        printMessage(f"protocol: HTTP", level=1)
+
+    # アドレス表示
+    if args.https == 1:
+        printMessage(f"open https://<IP>:<PORT>/ with your browser.", level=0)
+    else:
+        printMessage(f"open http://<IP>:<PORT>/ with your browser.", level=0)
+        
+    if EX_PORT and EX_IP and args.https == 1:
+        printMessage(f"In many cases it is one of the following", level=1)
+        printMessage(f"https://localhost:{EX_PORT}/", level=1)
+        for ip in EX_IP.strip().split(" "):
+            printMessage(f"https://{ip}:{EX_PORT}/", level=1)
+    elif EX_PORT and EX_IP and args.https == 0:
+        printMessage(f"In many cases it is one of the following", level=1)
+        printMessage(f"http://localhost:{EX_PORT}/", level=1)
+
+
+    # サーバ起動
+    if args.https:
+        # HTTPS サーバ起動 
+        uvicorn.run(
+            f"{os.path.basename(__file__)[:-3]}:app_socketio", 
+            host="0.0.0.0", 
+            port=int(PORT), 
+            reload=True, 
+            ssl_keyfile = key_path,
+            ssl_certfile = cert_path,
+            log_level="critical"            
+        )
+    else:
+        # HTTP サーバ起動
+        uvicorn.run(
+            f"{os.path.basename(__file__)[:-3]}:app_socketio", 
+            host="0.0.0.0", 
+            port=int(PORT), 
+            reload=True,
+            log_level="critical"
+        )
+

demo/mods/VoiceChanger.py

@@ -0,0 +1,76 @@
+import torch
+from scipy.io.wavfile import write, read
+import numpy as np
+import struct, traceback
+
+import utils 
+import commons 
+from models import SynthesizerTrn
+from text.symbols import symbols
+from data_utils import TextAudioSpeakerLoader, TextAudioSpeakerCollate
+from mel_processing import spectrogram_torch
+from text import text_to_sequence, cleaned_text_to_sequence
+
+
+class VoiceChanger():
+    def __init__(self, config, model):
+        self.hps = utils.get_hparams_from_file(config)
+        self.net_g = SynthesizerTrn(
+                len(symbols),
+                self.hps.data.filter_length // 2 + 1,
+                self.hps.train.segment_size // self.hps.data.hop_length,
+                n_speakers=self.hps.data.n_speakers,
+                **self.hps.model)
+        self.net_g.eval()
+        self.gpu_num = torch.cuda.device_count()
+        utils.load_checkpoint( model, self.net_g, None)
+        print(f"VoiceChanger Initialized (GPU_NUM:{self.gpu_num})")
+
+    def destroy(self):
+        del self.net_g
+
+    def on_request(self, gpu, srcId, dstId, timestamp, wav): 
+        # if wav==0:
+        #     samplerate, data=read("dummy.wav")
+        #     unpackedData = data
+        # else:
+        #     unpackedData = np.array(struct.unpack('<%sh'%(len(wav) // struct.calcsize('<h') ), wav))
+        #     write("logs/received_data.wav", 24000, unpackedData.astype(np.int16))
+
+        unpackedData = wav
+
+        try:
+
+            text_norm = text_to_sequence("a", self.hps.data.text_cleaners)
+            text_norm = commons.intersperse(text_norm, 0)
+            text_norm = torch.LongTensor(text_norm)
+
+            audio = torch.FloatTensor(unpackedData.astype(np.float32))
+            audio_norm = audio /self.hps.data.max_wav_value
+            audio_norm = audio_norm.unsqueeze(0)
+
+            spec = spectrogram_torch(audio_norm, self.hps.data.filter_length,
+                    self.hps.data.sampling_rate, self.hps.data.hop_length, self.hps.data.win_length,
+                    center=False)
+            spec = torch.squeeze(spec, 0)
+            sid = torch.LongTensor([int(srcId)])
+            
+            data =  (text_norm, spec, audio_norm, sid)
+            data = TextAudioSpeakerCollate()([data])
+
+            if gpu<0 or self.gpu_num==0 :
+                with torch.no_grad():
+                    x, x_lengths, spec, spec_lengths, y, y_lengths, sid_src = [x.cpu() for x in data]
+                    sid_tgt1 = torch.LongTensor([dstId]).cpu()
+                    audio1 = (self.net_g.cpu().voice_conversion(spec, spec_lengths, sid_src=sid_src, sid_tgt=sid_tgt1)[0][0,0].data * self.hps.data.max_wav_value).cpu().float().numpy()
+            else:
+                with torch.no_grad():
+                    x, x_lengths, spec, spec_lengths, y, y_lengths, sid_src = [x.cuda(gpu) for x in data]
+                    sid_tgt1 = torch.LongTensor([dstId]).cuda(gpu)
+                    audio1 = (self.net_g.cuda(gpu).voice_conversion(spec, spec_lengths, sid_src=sid_src, sid_tgt=sid_tgt1)[0][0,0].data * self.hps.data.max_wav_value).cpu().float().numpy()
+        except Exception as e:
+            print("VC PROCESSING!!!! EXCEPTION!!!", e)
+            print(traceback.format_exc())
+        
+        audio1 = audio1.astype(np.int16)
+        return audio1

demo/mods/ssl.py

@@ -0,0 +1,24 @@
+import os
+from OpenSSL import crypto
+
+def create_self_signed_cert(certfile, keyfile, certargs, cert_dir="."):
+    C_F = os.path.join(cert_dir, certfile)
+    K_F = os.path.join(cert_dir, keyfile)
+    if not os.path.exists(C_F) or not os.path.exists(K_F):
+        k = crypto.PKey()
+        k.generate_key(crypto.TYPE_RSA, 2048)
+        cert = crypto.X509()
+        cert.get_subject().C = certargs["Country"]
+        cert.get_subject().ST = certargs["State"]
+        cert.get_subject().L = certargs["City"]
+        cert.get_subject().O = certargs["Organization"]
+        cert.get_subject().OU = certargs["Org. Unit"]
+        cert.get_subject().CN = 'Example'
+        cert.set_serial_number(1000)
+        cert.gmtime_adj_notBefore(0)
+        cert.gmtime_adj_notAfter(315360000)
+        cert.set_issuer(cert.get_subject())
+        cert.set_pubkey(k)
+        cert.sign(k, 'sha1')
+        open(C_F, "wb").write(crypto.dump_certificate(crypto.FILETYPE_PEM, cert))
+        open(K_F, "wb").write(crypto.dump_privatekey(crypto.FILETYPE_PEM, k))

demo/serverSIO.py

@@ -22,7 +22,12 @@ from mel_processing import spectrogram_torch
 from text import text_to_sequence, cleaned_text_to_sequence
 
 class MyCustomNamespace(socketio.Namespace): 
-    def __init__(self, namespace, config, model):
+    def __init__(self, namespace):
+        super().__init__(namespace)
+        self.gpu_num = torch.cuda.device_count()
+        print("GPU_NUM:",self.gpu_num)
+
+    def __init__old(self, namespace, config, model):
         super().__init__(namespace)
         self.hps =utils.get_hparams_from_file(config)
         self.net_g = SynthesizerTrn(
@@ -36,12 +41,37 @@ class MyCustomNamespace(socketio.Namespace):
         print("GPU_NUM:",self.gpu_num)
         utils.load_checkpoint( model, self.net_g, None)
 
+    def loadModel(self, config, model):
+        self.hps =utils.get_hparams_from_file(config)
+        print("before DELETE:", torch.cuda.memory_allocated())
+        if hasattr(self, 'net_g') == True:
+            print("DELETE MODEL:", torch.cuda.memory_allocated())
+            del self.net_g
+        print("before load", torch.cuda.memory_allocated())
+        self.net_g = SynthesizerTrn(
+                len(symbols),
+                self.hps.data.filter_length // 2 + 1,
+                self.hps.train.segment_size // self.hps.data.hop_length,
+                n_speakers=self.hps.data.n_speakers,
+                **self.hps.model)
+        self.net_g.eval()
+        utils.load_checkpoint( model, self.net_g, None)
+        print(torch.cuda.memory_allocated())
+        print("after load", torch.cuda.memory_allocated())
+
+
+
     def on_connect(self, sid, environ):
         print('[{}] connet sid : {}'.format(datetime.now().strftime('%Y-%m-%d %H:%M:%S') , sid))
         # print('[{}] connet env : {}'.format(datetime.now().strftime('%Y-%m-%d %H:%M:%S') , environ))
 
+    def on_load_model(self, sid, msg):
+        print("on_load_model")
+        print(msg)
+        pass
+
     def on_request_message(self, sid, msg): 
-        # print("MESSGaa", msg)
+        print("on_request_message", torch.cuda.memory_allocated())
         gpu = int(msg[0])
         srcId = int(msg[1])
         dstId = int(msg[2])
@@ -223,7 +253,17 @@ if __name__ == '__main__':
 
     # SocketIOセットアップ
     sio = socketio.Server(cors_allowed_origins='*') 
-    sio.register_namespace(MyCustomNamespace('/test', CONFIG, MODEL)) 
+    namespace = MyCustomNamespace('/test')
+    sio.register_namespace(namespace) 
+    print("loadmodel1:")
+    namespace.loadModel(CONFIG, MODEL)
+    print("loadmodel2:")
+    namespace.loadModel(CONFIG, MODEL)
+    print("loadmodel3:")
+    namespace.loadModel(CONFIG, MODEL)
+    print("loadmodel4:")
+    namespace.loadModel(CONFIG, MODEL)
+    print("loadmodel5:")
     app = socketio.WSGIApp(sio,static_files={
         '': '../frontend/dist',
         '/': '../frontend/dist/index.html',

demo/setup.sh

@@ -12,32 +12,17 @@ if [[ -e ./setting.json ]]; then
   echo "カスタムセッティングを使用"
   cp ./setting.json ../frontend/dist/assets/setting.json
 else
-  if [ "${TYPE}" = "SOFT_VC" ] ; then
+  cp ../frontend/dist/assets/setting_mmvc.json ../frontend/dist/assets/setting.json  
-    cp ../frontend/dist/assets/setting_softvc.json ../frontend/dist/assets/setting.json
-  elif [ "${TYPE}" = "SOFT_VC_FAST_API" ] ; then
-    cp ../frontend/dist/assets/setting_softvc_colab.json ../frontend/dist/assets/setting.json
-  else
-    cp ../frontend/dist/assets/setting_mmvc.json ../frontend/dist/assets/setting.json  
-  fi
 fi
 
 
 # 起動
-if [ "${TYPE}" = "SOFT_VC" ] ; then
+if [ "${TYPE}" = "MMVC" ] ; then
-  echo "SOFT_VCを起動します"
-  python3 SoftVcServerSIO.py $PARAMS 2>stderr.txt
-elif [ "${TYPE}" = "SOFT_VC_VERBOSE" ] ; then
-  echo "SOFT_VCを起動します(verbose)"
-  python3 SoftVcServerSIO.py $PARAMS 
-elif [ "${TYPE}" = "SOFT_VC_FAST_API" ] ; then
-  echo "SOFT_VC_FAST_APIを起動します"
-  python3 SoftVcServerFastAPI.py 8080 docker
-elif [ "${TYPE}" = "MMVC" ] ; then
   echo "MMVCを起動します"
-  python3 serverSIO.py $PARAMS 2>stderr.txt
+  python3 MMVCServerSIO.py $PARAMS 2>stderr.txt
 elif [ "${TYPE}" = "MMVC_VERBOSE" ] ; then
   echo "MMVCを起動します(verbose)"
-  python3 serverSIO.py $PARAMS
+  python3 MMVCServerSIO.py $PARAMS
 fi
 
 

start2.sh

@@ -1,7 +1,7 @@
 #!/bin/bash
 set -eu
 
-DOCKER_IMAGE=dannadori/voice-changer:20221028_220714
+DOCKER_IMAGE=dannadori/voice-changer:20221029_233016
 #DOCKER_IMAGE=voice-changer
 
 

start_v0.1.sh

@@ -0,0 +1,314 @@
+#!/bin/bash
+set -eu
+
+DOCKER_IMAGE=dannadori/voice-changer:20221028_220714
+#DOCKER_IMAGE=voice-changer
+
+
+MODE=$1
+PARAMS=${@:2:($#-1)}
+
+### DEFAULT VAR ###
+DEFAULT_EX_PORT=18888
+DEFAULT_USE_GPU=on # on|off
+DEFAULT_VERBOSE=off # on|off
+
+### ENV VAR ###
+EX_PORT=${EX_PORT:-${DEFAULT_EX_PORT}}
+USE_GPU=${USE_GPU:-${DEFAULT_USE_GPU}}
+VERBOSE=${VERBOSE:-${DEFAULT_VERBOSE}}
+
+#echo $EX_PORT $USE_GPU $VERBOSE
+
+### INTERNAL SETTING ###
+TENSORBOARD_PORT=6006
+SIO_PORT=8080
+
+
+### 
+if [ "${MODE}" = "MMVC_TRAIN" ]; then
+    echo "トレーニングを開始します"
+
+    docker run -it --gpus all --shm-size=128M \
+        -v `pwd`/exp/${name}/dataset:/MMVC_Trainer/dataset \
+        -v `pwd`/exp/${name}/logs:/MMVC_Trainer/logs \
+        -v `pwd`/exp/${name}/filelists:/MMVC_Trainer/filelists \
+        -v `pwd`/vc_resources:/resources \
+        -e LOCAL_UID=$(id -u $USER) \
+        -e LOCAL_GID=$(id -g $USER) \
+        -e EX_IP="`hostname -I`" \
+        -e EX_PORT=${EX_PORT} \
+        -e VERBOSE=${VERBOSE} \        
+        -p ${EX_PORT}:6006 $DOCKER_IMAGE "$@"
+
+elif [ "${MODE}" = "MMVC" ]; then
+    if [ "${USE_GPU}" = "on" ]; then
+        echo "MMVCを起動します(with gpu)"
+
+        docker run -it --gpus all --shm-size=128M \
+        -v `pwd`/vc_resources:/resources \
+        -e LOCAL_UID=$(id -u $USER) \
+        -e LOCAL_GID=$(id -g $USER) \
+        -e EX_IP="`hostname -I`" \
+        -e EX_PORT=${EX_PORT} \
+        -e VERBOSE=${VERBOSE} \
+        -p ${EX_PORT}:8080 $DOCKER_IMAGE "$@"
+    else
+        echo "MMVCを起動します(only cpu)"
+        docker run -it --shm-size=128M \
+        -v `pwd`/vc_resources:/resources \
+        -e LOCAL_UID=$(id -u $USER) \
+        -e LOCAL_GID=$(id -g $USER) \
+        -e EX_IP="`hostname -I`" \
+        -e EX_PORT=${EX_PORT} \
+        -e VERBOSE=${VERBOSE} \
+        -p ${EX_PORT}:8080 $DOCKER_IMAGE "$@"
+
+        # docker run -it --shm-size=128M \
+        # -v `pwd`/vc_resources:/resources \
+        # -e LOCAL_UID=$(id -u $USER) \
+        # -e LOCAL_GID=$(id -g $USER) \
+        # -e EX_IP="`hostname -I`" \
+        # -e EX_PORT=${EX_PORT} \
+        # -e VERBOSE=${VERBOSE} \
+        # --entrypoint="" \
+        # -p ${EX_PORT}:8080 $DOCKER_IMAGE /bin/bash
+
+    fi
+
+elif [ "${MODE}" = "SOFT_VC" ]; then
+    if [ "${USE_GPU}" = "on" ]; then
+        echo "Start Soft-vc"
+
+        docker run -it --gpus all --shm-size=128M \
+        -v `pwd`/vc_resources:/resources \
+        -e LOCAL_UID=$(id -u $USER) \
+        -e LOCAL_GID=$(id -g $USER) \
+        -e EX_IP="`hostname -I`" \
+        -e EX_PORT=${EX_PORT} \
+        -e VERBOSE=${VERBOSE} \
+        -p ${EX_PORT}:8080 $DOCKER_IMAGE "$@"
+    else
+        echo "Start Soft-vc withou GPU is not supported"
+    fi
+
+else
+    echo "
+usage: 
+    $0 <MODE> <params...>
+    MODE: select one of ['MMVC_TRAIN', 'MMVC', 'SOFT_VC']
+" >&2
+fi
+
+
+
+
+# echo $EX_PORT
+
+
+# echo "------"
+# echo "$@"
+# echo "------"
+
+# # usage() {
+# #     echo "
+# # usage: 
+# #     For training
+# #         $0 [-t] -n <exp_name> [-b batch_size] [-r] 
+# #             -t: トレーニングモードで実行する場合に指定してください。(train)
+# #             -n: トレーニングの名前です。(name)
+# #             -b: バッチサイズです。(batchsize)
+# #             -r: トレーニング再開の場合に指定してください。(resume)
+# #     For changing voice
+# #         $0 [-v] [-c config] [-m model] [-g on/off]
+# #             -v: ボイスチェンジャーモードで実行する場合に指定してください。(voice changer)
+# #             -c: トレーニングで使用したConfigのファイル名です。(config)
+# #             -m: トレーニング済みのモデルのファイル名です。(model)
+# #             -g: GPU使用/不使用。デフォルトはonなのでGPUを使う場合は指定不要。(gpu)
+# #             -p: port番号
+# #     For help
+# #         $0 [-h]
+# #             -h: show this help
+# # " >&2
+# # }
+# # warn () {
+# #     echo "! ! ! $1 ! ! !"
+# #     exit 1
+# # }
+
+
+# # training_flag=false
+# # name=999_exp
+# # batch_size=10
+# # resume_flag=false
+
+# # voice_change_flag=false
+# # config=
+# # model=
+# # gpu=on
+# # port=8080
+# # escape_flag=false
+
+# # # オプション解析
+# # while getopts tn:b:rvc:m:g:p:hx OPT; do
+# #     case $OPT in
+# #     t) 
+# #         training_flag=true
+# #         ;;
+# #     n) 
+# #         name="$OPTARG"
+# #         ;;
+# #     b) 
+# #         batch_size="$OPTARG"
+# #         ;;
+# #     r) 
+# #         resume_flag=true
+# #         ;;
+# #     v) 
+# #         voice_change_flag=true
+# #         ;;
+# #     c) 
+# #         config="$OPTARG"
+# #         ;;
+# #     m) 
+# #         model="$OPTARG"
+# #         ;;
+# #     g) 
+# #         gpu="$OPTARG"
+# #         ;;
+# #     p) 
+# #         port="$OPTARG"
+# #         ;;
+# #     h | \?) 
+# #         usage && exit 1
+# #         ;;
+# #     x)
+# #         escape_flag=true
+# #     esac
+# # done
+
+
+# # # モード解析
+# # if $training_flag && $voice_change_flag; then
+# #     warn "-t（トレーニングモード） と -v（ボイチェンモード）は同時に指定できません。"
+# # elif $training_flag; then
+# #     echo "■■■  ト レ ー ニ ン グ モ ー ド   ■■■"
+# # elif $voice_change_flag; then
+# #     echo "■■■  ボ イ チ ェ ン モ ー ド  ■■■"
+# # elif $escape_flag; then
+# #     /bin/bash
+# # else
+# #     warn "-t（トレーニングモード） と -v（ボイチェンモード）のいずれかを指定してください。"
+# # fi
+
+# if [ "${MODE}" = "MMVC_TRAIN_INITIAL" ]; then
+#     echo "トレーニングを開始します"
+# elif [ "${MODE}" = "MMVC" ]; then
+#     echo "MMVCを起動します"
+
+#     docker run -it --gpus all --shm-size=128M \
+#     -v `pwd`/vc_resources:/resources \
+#     -e LOCAL_UID=$(id -u $USER) \
+#     -e LOCAL_GID=$(id -g $USER) \
+#     -e EX_IP="`hostname -I`" \
+#     -e EX_PORT=${port} \
+#     -p ${port}:8080 $DOCKER_IMAGE -v -c ${config} -m ${model}
+
+# elif [ "${MODE}" = "MMVC_VERBOSE" ]; then
+#     echo "MMVCを起動します(verbose)"
+# elif [ "${MODE}" = "MMVC_CPU" ]; then
+#     echo "MMVCを起動します(CPU)"
+# elif [ "${MODE}" = "MMVC_CPU_VERBOSE" ]; then
+#     echo "MMVCを起動します(CPU)(verbose)"
+# elif [ "${MODE}" = "SOFT_VC" ]; then
+#     echo "Start Soft-vc"
+# elif [ "${MODE}" = "SOFT_VC_VERBOSE" ]; then
+#     echo "Start Soft-vc(verbose)"
+# else
+#     echo "
+# usage: 
+#     $0 <MODE> <params...>
+#     EX_PORT: 
+#     MODE: one of ['MMVC_TRAIN', 'MMVC', 'SOFT_VC']
+
+#     For 'MMVC_TRAIN':
+#         $0 MMVC_TRAIN_INITIAL -n <exp_name> [-b batch_size] [-r] 
+#             -n: トレーニングの名前です。(name)
+#             -b: バッチサイズです。(batchsize)
+#             -r: トレーニング再開の場合に指定してください。(resume)
+#     For 'MMVC'
+#         $0 MMVC [-c config] [-m model] [-g on/off] [-p port] [-v]
+#             -c: トレーニングで使用したConfigのファイル名です。(config)
+#             -m: トレーニング済みのモデルのファイル名です。(model)
+#             -g: GPU使用/不使用。デフォルトはonなのでGPUを使う場合は指定不要。(gpu)
+#             -p: Docker からExposeするport番号
+#             -v: verbose
+#     For 'SOFT_VC'
+#         $0 SOFT_VC [-c config] [-m model] [-g on/off]
+#             -p: port exposed from docker container.
+#             -v: verbose
+# " >&2
+# fi
+
+
+
+# # if $training_flag; then
+# #     if $resume_flag; then
+# #         echo "トレーニングを再開します"
+# #         docker run -it --gpus all --shm-size=128M \
+# #             -v `pwd`/exp/${name}/dataset:/MMVC_Trainer/dataset \
+# #             -v `pwd`/exp/${name}/logs:/MMVC_Trainer/logs \
+# #             -v `pwd`/exp/${name}/filelists:/MMVC_Trainer/filelists \
+# #             -v `pwd`/vc_resources:/resources \
+# #             -e LOCAL_UID=$(id -u $USER) \
+# #             -e LOCAL_GID=$(id -g $USER) \
+# #             -p ${TENSORBOARD_PORT}:6006 $DOCKER_IMAGE -t -b ${batch_size} -r
+# #     else
+# #         echo "トレーニングを開始します"
+# #         docker run -it --gpus all --shm-size=128M \
+# #             -v `pwd`/exp/${name}/dataset:/MMVC_Trainer/dataset \
+# #             -v `pwd`/exp/${name}/logs:/MMVC_Trainer/logs \
+# #             -v `pwd`/exp/${name}/filelists:/MMVC_Trainer/filelists \
+# #             -v `pwd`/vc_resources:/resources \
+# #             -e LOCAL_UID=$(id -u $USER) \
+# #             -e LOCAL_GID=$(id -g $USER) \
+# #             -p ${TENSORBOARD_PORT}:6006 $DOCKER_IMAGE -t -b ${batch_size}
+# #     fi
+# # fi
+
+# # if $voice_change_flag; then
+# #     if [[ -z "$config" ]]; then
+# #         warn "コンフィグファイル(-c)を指定してください"
+# #     fi
+# #     if [[ -z "$model" ]]; then
+# #         warn "モデルファイル(-m)を指定してください"
+# #     fi
+# #     if [ "${gpu}" = "on" ]; then
+# #         echo "GPUをマウントして起動します。"
+
+# #         docker run -it --gpus all --shm-size=128M \
+# #         -v `pwd`/vc_resources:/resources \
+# #         -e LOCAL_UID=$(id -u $USER) \
+# #         -e LOCAL_GID=$(id -g $USER) \
+# #         -e EX_IP="`hostname -I`" \
+# #         -e EX_PORT=${port} \
+# #         -p ${port}:8080 $DOCKER_IMAGE -v -c ${config} -m ${model}
+# #     elif [ "${gpu}" = "off" ]; then
+# #         echo "CPUのみで稼働します。GPUは使用できません。"
+# #         docker run -it --shm-size=128M \
+# #         -v `pwd`/vc_resources:/resources \
+# #         -e LOCAL_UID=$(id -u $USER) \
+# #         -e LOCAL_GID=$(id -g $USER) \
+# #         -e EX_IP="`hostname -I`" \
+# #         -e EX_PORT=${port} \
+# #         -p ${port}:8080 $DOCKER_IMAGE -v -c ${config} -m ${model}
+# #     else
+# #         echo ${gpu}
+# #         warn "-g は onかoffで指定して下さい。"
+        
+# #     fi
+
+
+# # fi
+
+

trainer/Dockerfile

@@ -1,4 +1,4 @@
-FROM dannadori/voice-changer-internal:20221028_220538 as front
+FROM dannadori/voice-changer-internal:20221029_231527 as front
 FROM debian:bullseye-slim as base
 
 ARG DEBIAN_FRONTEND=noninteractive
@@ -8,7 +8,7 @@ RUN apt-get install -y python3-pip git
 RUN apt-get install -y espeak
 RUN apt-get install -y cmake
 
-RUN git clone --depth 1 https://github.com/isletennos/MMVC_Trainer.git -b v1.3.1.3
+#RUN git clone --depth 1 https://github.com/isletennos/MMVC_Trainer.git -b v1.3.1.3
 
 RUN pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu113
 
@@ -24,17 +24,20 @@ RUN pip install tqdm==4.64.0
 RUN pip install retry==0.9.2
 RUN pip install psutil==5.9.1
 RUN pip install python-socketio==5.7.1
-RUN pip install eventlet==0.33.1
+RUN pip install matplotlib==3.5.3
+
+RUN pip install fastapi==0.85.0
+RUN pip install python-multipart==0.0.5
+RUN pip install uvicorn==0.18.3
+RUN pip install websockets==10.4
+RUN pip install pyOpenSSL==22.0.0
 
 RUN pip install pyopenjtalk==0.2.0
 RUN pip install tensorboard==2.10.0
-RUN pip install matplotlib==3.5.3
 
-RUN pip install pyOpenSSL==22.0.0
+# WORKDIR /MMVC_Trainer/monotonic_align
-
+# RUN cythonize -3 -i core.pyx \
-WORKDIR /MMVC_Trainer/monotonic_align
+#  && mv core.cpython-39-x86_64-linux-gnu.so monotonic_align/
-RUN cythonize -3 -i core.pyx \
- && mv core.cpython-39-x86_64-linux-gnu.so monotonic_align/
 
 
 FROM debian:bullseye-slim
@@ -64,12 +67,11 @@ COPY --from=front --chmod=777 /voice-changer-internal/frontend/dist /voice-chang
 COPY --from=front --chmod=777 /voice-changer-internal/voice-change-service /voice-changer-internal/voice-change-service
 RUN chmod 0777 /voice-changer-internal/voice-change-service 
 
-##### Soft VC
+# ##### Soft VC
-COPY --from=front /hubert /hubert
+# COPY --from=front /hubert /hubert
-COPY --from=front /acoustic-model /acoustic-model
+# COPY --from=front /acoustic-model /acoustic-model
-COPY --from=front /hifigan /hifigan
+# COPY --from=front /hifigan /hifigan
-
+# COPY --from=front /models /models 
-COPY --from=front /models /models 
 
 
 ENTRYPOINT ["/bin/bash", "setup.sh"]