Add deterministic mode for testing (same text, same audio guaranteed)

mimic3-http/mimic3_http/VERSION

@@ -1 +1 @@
-0.1.0
+0.1.1

mimic3-http/mimic3_http/app.py

@@ -70,6 +70,11 @@ def get_app(args: argparse.Namespace, request_queue: Queue, temp_dir: str):
 
         Returns: WAV bytes
         """
+        if args.deterministic:
+            # Disable noise
+            _LOGGER.debug("Disabling noise in deterministic mode")
+            params.noise_scale = 0.0
+            params.noise_w = 0.0
 
         _LOGGER.debug(params)
 

mimic3-http/mimic3_http/args.py

@@ -70,6 +70,11 @@ def get_args(argv=None) -> argparse.Namespace:
         action="store_true",
         help="Use Onnx CUDA execution provider (requires onnxruntime-gpu)",
     )
+    parser.add_argument(
+        "--deterministic",
+        action="store_true",
+        help="Ensure that the same audio is always synthesized from the same text",
+    )
     parser.add_argument(
         "--num-threads",
         type=int,

mimic3-http/mimic3_http/synthesis.py

@@ -85,6 +85,7 @@ def do_synthesis_proc(args: argparse.Namespace, request_queue: Queue):
                 noise_w=args.noise_w,
                 use_cuda=args.cuda,
                 voices_directories=args.voices_dir,
+                use_deterministic_compute=args.deterministic,
             )
         )
 

mimic3-tts/mimic3_tts/VERSION

@@ -1 +1 @@
-0.1.6
+0.1.7

mimic3-tts/mimic3_tts/__main__.py

@@ -221,6 +221,12 @@ def initialize_args(state: CommandLineInterfaceState):
         # Split apart voice
         args.voice, args.speaker = args.voice.split("#", maxsplit=1)
 
+    if args.deterministic:
+        # Disable noise
+        _LOGGER.debug("Disabling noise in deterministic mode")
+        args.noise_scale = 0.0
+        args.noise_w = 0.0
+
 
 def initialize_tts(state: CommandLineInterfaceState):
     """Create Mimic 3 TTS from command-line arguments"""
@@ -232,9 +238,13 @@ def initialize_tts(state: CommandLineInterfaceState):
         # Local TTS
         state.tts = Mimic3TextToSpeechSystem(
             Mimic3Settings(
+                length_scale=args.length_scale,
+                noise_scale=args.noise_scale,
+                noise_w=args.noise_w,
                 voices_directories=args.voices_dir,
                 speaker=args.speaker,
                 use_cuda=args.cuda,
+                use_deterministic_compute=args.deterministic,
             )
         )
 
@@ -694,6 +704,11 @@ def get_args(argv=None):
         action="store_true",
         help="Use Onnx CUDA execution provider (requires onnxruntime-gpu)",
     )
+    parser.add_argument(
+        "--deterministic",
+        action="store_true",
+        help="Ensure that the same audio is always synthesized from the same text",
+    )
     parser.add_argument("--seed", type=int, help="Set random seed (default: not set)")
     parser.add_argument("--version", action="store_true", help="Print version and exit")
     parser.add_argument(

mimic3-tts/mimic3_tts/tts.py

@@ -117,6 +117,9 @@ class Mimic3Settings:
     rate: float = DEFAULT_RATE
     """Voice speaking rate (< 1 is slower, > 1 is faster)"""
 
+    use_deterministic_compute: bool = False
+    """Force onnxruntime to use deterministic compute mode. For fully deterministic synthesis, also set noise_scale and noise_w to 0."""
+
 
 @dataclass
 class Mimic3Phonemes:
@@ -543,6 +546,7 @@ class Mimic3TextToSpeechSystem(TextToSpeechSystem):
             model_dir,
             providers=providers,
             share_models=self.settings.share_onnx_models_between_threads,
+            use_deterministic_compute=self.settings.use_deterministic_compute,
         )
 
         _LOGGER.info("Loaded voice from %s", model_dir)

mimic3-tts/mimic3_tts/voice.py

@@ -252,6 +252,7 @@ class Mimic3Voice(metaclass=ABCMeta):
             ]
         ] = None,
         share_models: bool = True,
+        use_deterministic_compute: bool = False,
     ) -> "Mimic3Voice":
         """Load a Mimic 3 voice from a directory"""
         voice_dir = Path(voice_dir)
@@ -283,6 +284,7 @@ class Mimic3Voice(metaclass=ABCMeta):
                         generator_path,
                         session_options=session_options,
                         providers=providers,
+                        use_deterministic_compute=use_deterministic_compute,
                     )
 
                     Mimic3Voice._SHARED_MODELS[model_key] = onnx_model
@@ -293,6 +295,7 @@ class Mimic3Voice(metaclass=ABCMeta):
                 generator_path,
                 session_options=session_options,
                 providers=providers,
+                use_deterministic_compute=use_deterministic_compute,
             )
 
         # phoneme -> phoneme, phoneme, ...
@@ -381,6 +384,7 @@ class Mimic3Voice(metaclass=ABCMeta):
                 typing.Union[str, typing.Tuple[str, typing.Dict[str, typing.Any]]]
             ]
         ] = None,
+        use_deterministic_compute: bool = False,
     ) -> onnxruntime.InferenceSession:
         _LOGGER.debug("Loading model from %s", generator_path)
 
@@ -394,6 +398,8 @@ class Mimic3Voice(metaclass=ABCMeta):
                     onnxruntime.GraphOptimizationLevel.ORT_DISABLE_ALL
                 )
 
+        session_options.use_deterministic_compute = use_deterministic_compute
+
         onnx_model = onnxruntime.InferenceSession(
             str(generator_path), sess_options=session_options, providers=providers
         )