対応デバイスが分かるAPIエンドポイントを追加 (#299)

* /supported_devicesを追加

* model修正

* 抽象プロパティを定義

* コメント追加

run.py

@@ -31,6 +31,7 @@
     ParseKanaError,
     Speaker,
     SpeakerInfo,
+    SupportedDevicesInfo,
 )
 from voicevox_engine.morphing import synthesis_morphing
 from voicevox_engine.morphing import (
@@ -532,6 +533,18 @@ def speaker_info(speaker_uuid: str, core_version: Optional[str] = None):
         ret_data = {"policy": policy, "portrait": portrait, "style_infos": style_infos}
         return ret_data
 
+    @app.get("/supported_devices", response_model=SupportedDevicesInfo, tags=["その他"])
+    def supported_devices(
+        core_version: Optional[str] = None,
+    ):
+        supported_devices = get_engine(core_version).supported_devices
+        if supported_devices is None:
+            raise HTTPException(status_code=422, detail="非対応の機能です。")
+        return Response(
+            content=supported_devices,
+            media_type="application/json",
+        )
+
     return app
 
 

test/test_mock_synthesis_engine.py

@@ -102,7 +102,7 @@ def setUp(self):
                 pause_mora=None,
             ),
         ]
-        self.engine = MockSynthesisEngine(speakers="")
+        self.engine = MockSynthesisEngine(speakers="", supported_devices="")
 
     def test_replace_phoneme_length(self):
         self.assertEqual(

voicevox_engine/dev/core/__init__.py

@@ -2,6 +2,7 @@
     decode_forward,
     initialize,
     metas,
+    supported_devices,
     yukarin_s_forward,
     yukarin_sa_forward,
 )
@@ -12,4 +13,5 @@
     "yukarin_s_forward",
     "yukarin_sa_forward",
     "metas",
+    "supported_devices",
 ]

voicevox_engine/dev/core/mock.py

@@ -110,3 +110,12 @@ def metas() -> str:
             },
         ]
     )
+
+
+def supported_devices() -> str:
+    return json.dumps(
+        {
+            "cpu": True,
+            "cuda": False,
+        }
+    )

voicevox_engine/dev/synthesis_engine/mock.py

@@ -1,5 +1,5 @@
 from logging import getLogger
-from typing import Any, Dict, List
+from typing import Any, Dict, List, Optional
 
 import numpy as np
 from pyopenjtalk import tts
@@ -21,9 +21,18 @@ def __init__(self, **kwargs):
         """
         super().__init__()
 
-        self.speakers = kwargs["speakers"]
+        self._speakers = kwargs["speakers"]
+        self._supported_devices = kwargs["supported_devices"]
         self.default_sampling_rate = 24000
 
+    @property
+    def speakers(self) -> str:
+        return self._speakers
+
+    @property
+    def supported_devices(self) -> Optional[str]:
+        return self._supported_devices
+
     def replace_phoneme_length(
         self, accent_phrases: List[AccentPhrase], speaker_id: int
     ) -> List[AccentPhrase]:

voicevox_engine/model.py

@@ -140,3 +140,12 @@ class SpeakerInfo(BaseModel):
     policy: str = Field(title="policy.md")
     portrait: str = Field(title="portrait.pngをbase64エンコードしたもの")
     style_infos: List[StyleInfo] = Field(title="スタイルの追加情報")
+
+
+class SupportedDevicesInfo(BaseModel):
+    """
+    対応しているデバイスの情報
+    """
+
+    cpu: bool = Field(title="CPUに対応しているか")
+    cuda: bool = Field(title="CUDA(GPU)に対応しているか")

voicevox_engine/synthesis_engine/make_synthesis_engines.py

@@ -81,11 +81,18 @@ def make_synthesis_engines(
                     file=sys.stderr,
                 )
                 continue
+            try:
+                supported_devices = core.supported_devices()
+            except NameError:
+                # libtorch版コアは対応していないのでNameErrorになる
+                # 対応デバイスが不明であることを示すNoneを代入する
+                supported_devices = None
             synthesis_engines[core_version] = SynthesisEngine(
                 yukarin_s_forwarder=core.yukarin_s_forward,
                 yukarin_sa_forwarder=core.yukarin_sa_forward,
                 decode_forwarder=core.decode_forward,
                 speakers=core.metas(),
+                supported_devices=supported_devices,
             )
         except Exception:
             if not enable_mock:
@@ -96,9 +103,12 @@ def make_synthesis_engines(
                 file=sys.stderr,
             )
             from ..dev.core import metas as mock_metas
+            from ..dev.core import supported_devices as mock_supported_devices
             from ..dev.synthesis_engine import MockSynthesisEngine
 
             if "0.0.0" not in synthesis_engines:
-                synthesis_engines["0.0.0"] = MockSynthesisEngine(speakers=mock_metas())
+                synthesis_engines["0.0.0"] = MockSynthesisEngine(
+                    speakers=mock_metas(), supported_devices=mock_supported_devices()
+                )
 
     return synthesis_engines

voicevox_engine/synthesis_engine/synthesis_engine.py

@@ -132,6 +132,7 @@ def __init__(
         yukarin_sa_forwarder,
         decode_forwarder,
         speakers: str,
+        supported_devices: Optional[str] = None,
     ):
         """
         yukarin_s_forwarder: 音素列から、音素ごとの長さを求める関数
@@ -160,15 +161,28 @@ def __init__(
             return: 音声波形
 
         speakers: coreから取得したspeakersに関するjsonデータの文字列
+
+        supported_devices:
+            coreから取得した対応デバイスに関するjsonデータの文字列
+            Noneの場合はコアが情報の取得に対応していないため、対応デバイスは不明
         """
         super().__init__()
         self.yukarin_s_forwarder = yukarin_s_forwarder
         self.yukarin_sa_forwarder = yukarin_sa_forwarder
         self.decode_forwarder = decode_forwarder
 
-        self.speakers = speakers
+        self._speakers = speakers
+        self._supported_devices = supported_devices
         self.default_sampling_rate = 24000
 
+    @property
+    def speakers(self) -> str:
+        return self._speakers
+
+    @property
+    def supported_devices(self) -> Optional[str]:
+        return self._supported_devices
+
     def replace_phoneme_length(
         self, accent_phrases: List[AccentPhrase], speaker_id: int
     ) -> List[AccentPhrase]:

voicevox_engine/synthesis_engine/synthesis_engine_base.py

@@ -1,6 +1,6 @@
 import copy
 from abc import ABCMeta, abstractmethod
-from typing import List
+from typing import List, Optional
 
 from .. import full_context_label
 from ..full_context_label import extract_full_context_label
@@ -78,6 +78,16 @@ def full_context_label_moras_to_moras(
 
 
 class SynthesisEngineBase(metaclass=ABCMeta):
+    @property
+    @abstractmethod
+    def speakers(self) -> str:
+        raise NotImplementedError
+
+    @property
+    @abstractmethod
+    def supported_devices(self) -> Optional[str]:
+        raise NotImplementedError
+
     @abstractmethod
     def replace_phoneme_length(
         self, accent_phrases: List[AccentPhrase], speaker_id: int