Merge pull request #53 from huggingface/add-paraformer

Add paraformer - Chinese STT

STT/paraformer_handler.py

@@ -0,0 +1,61 @@
+import logging
+from time import perf_counter
+
+from baseHandler import BaseHandler
+from funasr import AutoModel
+import numpy as np
+from rich.console import Console
+import torch
+
+logging.basicConfig(
+    format="%(asctime)s - %(name)s - %(levelname)s - %(message)s",
+)
+logger = logging.getLogger(__name__)
+
+console = Console()
+
+
+class ParaformerSTTHandler(BaseHandler):
+    """
+    Handles the Speech To Text generation using a Paraformer model.
+    The default for this model is set to Chinese.
+    This model was contributed by @wuhongsheng.
+    """
+
+    def setup(
+        self,
+        model_name="paraformer-zh",
+        device="cuda",
+        gen_kwargs={},
+    ):
+        print(model_name)
+        if len(model_name.split("/")) > 1:
+            model_name = model_name.split("/")[-1]
+        self.device = device
+        self.model = AutoModel(model=model_name, device=device)
+        self.warmup()
+
+    def warmup(self):
+        logger.info(f"Warming up {self.__class__.__name__}")
+
+        # 2 warmup steps for no compile or compile mode with CUDA graphs capture
+        n_steps = 1
+        dummy_input = np.array([0] * 512, dtype=np.float32)
+        for _ in range(n_steps):
+            _ = self.model.generate(dummy_input)[0]["text"].strip().replace(" ", "")
+
+    def process(self, spoken_prompt):
+        logger.debug("infering paraformer...")
+
+        global pipeline_start
+        pipeline_start = perf_counter()
+
+        pred_text = (
+            self.model.generate(spoken_prompt)[0]["text"].strip().replace(" ", "")
+        )
+        torch.mps.empty_cache()
+
+        logger.debug("finished paraformer inference")
+        console.print(f"[yellow]USER: {pred_text}")
+
+        yield pred_text

arguments_classes/module_arguments.py

@@ -23,7 +23,7 @@ class ModuleArguments:
     stt: Optional[str] = field(
         default="whisper",
         metadata={
-            "help": "The STT to use. Either 'whisper' or 'whisper-mlx'. Default is 'whisper'."
+            "help": "The STT to use. Either 'whisper', 'whisper-mlx', and 'paraformer'. Default is 'whisper'."
         },
     )
     llm: Optional[str] = field(

arguments_classes/paraformer_stt_arguments.py

@@ -0,0 +1,17 @@
+from dataclasses import dataclass, field
+
+
+@dataclass
+class ParaformerSTTHandlerArguments:
+    paraformer_stt_model_name: str = field(
+        default="paraformer-zh",
+        metadata={
+            "help": "The pretrained model to use. Default is 'paraformer-zh'. Can be choose from https://github.com/modelscope/FunASR"
+        },
+    )
+    paraformer_stt_device: str = field(
+        default="cuda",
+        metadata={
+            "help": "The device type on which the model will run. Default is 'cuda' for GPU acceleration."
+        },
+    )

requirements.txt

@@ -2,4 +2,6 @@ nltk==3.9.1
 parler_tts @ git+https://github.com/huggingface/parler-tts.git
 melotts @ git+https://github.com/andimarafioti/MeloTTS.git#egg=MeloTTS  # made a copy of MeloTTS to have compatible versions of transformers
 torch==2.4.0
-sounddevice==0.5.0
+sounddevice==0.5.0
+funasr
+modelscope

requirements_mac.txt

@@ -4,4 +4,6 @@ melotts @ git+https://github.com/andimarafioti/MeloTTS.git#egg=MeloTTS  # made a
 torch==2.4.0
 sounddevice==0.5.0
 lightning-whisper-mlx>=0.0.10
-mlx-lm>=0.14.0
+mlx-lm>=0.14.0
+funasr>=1.1.6
+modelscope>=1.17.1

s2s_pipeline.py

@@ -13,6 +13,7 @@ from arguments_classes.mlx_language_model_arguments import (
     MLXLanguageModelHandlerArguments,
 )
 from arguments_classes.module_arguments import ModuleArguments
+from arguments_classes.paraformer_stt_arguments import ParaformerSTTHandlerArguments
 from arguments_classes.parler_tts_arguments import ParlerTTSHandlerArguments
 from arguments_classes.socket_receiver_arguments import SocketReceiverArguments
 from arguments_classes.socket_sender_arguments import SocketSenderArguments
@@ -73,6 +74,7 @@ def main():
             SocketSenderArguments,
             VADHandlerArguments,
             WhisperSTTHandlerArguments,
+            ParaformerSTTHandlerArguments,
             LanguageModelHandlerArguments,
             MLXLanguageModelHandlerArguments,
             ParlerTTSHandlerArguments,
@@ -89,6 +91,7 @@ def main():
             socket_sender_kwargs,
             vad_handler_kwargs,
             whisper_stt_handler_kwargs,
+            paraformer_stt_handler_kwargs,
             language_model_handler_kwargs,
             mlx_language_model_handler_kwargs,
             parler_tts_handler_kwargs,
@@ -102,6 +105,7 @@ def main():
             socket_sender_kwargs,
             vad_handler_kwargs,
             whisper_stt_handler_kwargs,
+            paraformer_stt_handler_kwargs,
             language_model_handler_kwargs,
             mlx_language_model_handler_kwargs,
             parler_tts_handler_kwargs,
@@ -163,6 +167,8 @@ def main():
                     kwargs.tts_device = common_device
                 if hasattr(kwargs, "stt_device"):
                     kwargs.stt_device = common_device
+                if hasattr(kwargs, "paraformer_stt_device"):
+                    kwargs.paraformer_stt_device = common_device
 
     # Call this function with the common device and all the handlers
     overwrite_device_argument(
@@ -171,9 +177,11 @@ def main():
         mlx_language_model_handler_kwargs,
         parler_tts_handler_kwargs,
         whisper_stt_handler_kwargs,
+        paraformer_stt_handler_kwargs,
     )
 
     prepare_args(whisper_stt_handler_kwargs, "stt")
+    prepare_args(paraformer_stt_handler_kwargs, "paraformer_stt")
     prepare_args(language_model_handler_kwargs, "lm")
     prepare_args(mlx_language_model_handler_kwargs, "mlx_lm")
     prepare_args(parler_tts_handler_kwargs, "tts")
@@ -243,8 +251,19 @@ def main():
             queue_out=text_prompt_queue,
             setup_kwargs=vars(whisper_stt_handler_kwargs),
         )
+    elif module_kwargs.stt == "paraformer":
+        from STT.paraformer_handler import ParaformerSTTHandler
+
+        stt = ParaformerSTTHandler(
+            stop_event,
+            queue_in=spoken_prompt_queue,
+            queue_out=text_prompt_queue,
+            setup_kwargs=vars(paraformer_stt_handler_kwargs),
+        )
     else:
-        raise ValueError("The STT should be either whisper or whisper-mlx")
+        raise ValueError(
+            "The STT should be either whisper, whisper-mlx, or paraformer."
+        )
     if module_kwargs.llm == "transformers":
         from LLM.language_model import LanguageModelHandler