init ascend tts

ascend_910-gpt-sovits/GPT-SoVITS/tools/asr/config.py

@@ -0,0 +1,44 @@
+import os
+
+
+def check_fw_local_models():
+    """
+    启动时检查本地是否有 Faster Whisper 模型.
+    """
+    model_size_list = [
+        "medium",
+        "medium.en",
+        "distil-large-v2",
+        "distil-large-v3",
+        "large-v1",
+        "large-v2",
+        "large-v3",
+    ]
+    for i, size in enumerate(model_size_list):
+        if os.path.exists(f"tools/asr/models/faster-whisper-{size}"):
+            model_size_list[i] = size + "-local"
+    return model_size_list
+
+
+def get_models():
+    model_size_list = [
+        "medium",
+        "medium.en",
+        "distil-large-v2",
+        "distil-large-v3",
+        "large-v1",
+        "large-v2",
+        "large-v3",
+    ]
+    return model_size_list
+
+
+asr_dict = {
+    "达摩 ASR (中文)": {"lang": ["zh", "yue"], "size": ["large"], "path": "funasr_asr.py", "precision": ["float32"]},
+    "Faster Whisper (多语种)": {
+        "lang": ["auto", "zh", "en", "ja", "ko", "yue"],
+        "size": get_models(),
+        "path": "fasterwhisper_asr.py",
+        "precision": ["float32", "float16", "int8"],
+    },
+}

ascend_910-gpt-sovits/GPT-SoVITS/tools/asr/fasterwhisper_asr.py

@@ -0,0 +1,170 @@
+import argparse
+import os
+import time
+import traceback
+
+import torch
+from faster_whisper import WhisperModel
+from huggingface_hub import snapshot_download
+from huggingface_hub.errors import LocalEntryNotFoundError
+from tqdm import tqdm
+
+from tools.asr.config import get_models
+from tools.asr.funasr_asr import only_asr
+from tools.my_utils import load_cudnn
+
+# fmt: off
+language_code_list = [
+    "af", "am", "ar", "as", "az", 
+    "ba", "be", "bg", "bn", "bo", 
+    "br", "bs", "ca", "cs", "cy", 
+    "da", "de", "el", "en", "es", 
+    "et", "eu", "fa", "fi", "fo", 
+    "fr", "gl", "gu", "ha", "haw", 
+    "he", "hi", "hr", "ht", "hu", 
+    "hy", "id", "is", "it", "ja", 
+    "jw", "ka", "kk", "km", "kn", 
+    "ko", "la", "lb", "ln", "lo", 
+    "lt", "lv", "mg", "mi", "mk", 
+    "ml", "mn", "mr", "ms", "mt", 
+    "my", "ne", "nl", "nn", "no", 
+    "oc", "pa", "pl", "ps", "pt", 
+    "ro", "ru", "sa", "sd", "si", 
+    "sk", "sl", "sn", "so", "sq", 
+    "sr", "su", "sv", "sw", "ta", 
+    "te", "tg", "th", "tk", "tl", 
+    "tr", "tt", "uk", "ur", "uz", 
+    "vi", "yi", "yo", "zh", "yue",
+    "auto"] 
+# fmt: on
+
+
+def download_model(model_size: str):
+    if "distil" in model_size:
+        repo_id = "Systran/faster-{}-whisper-{}".format(*model_size.split("-", maxsplit=1))
+    else:
+        repo_id = f"Systran/faster-whisper-{model_size}"
+    model_path = f"tools/asr/models/{repo_id.strip('Systran/')}"
+
+    files: list[str] = [
+        "config.json",
+        "model.bin",
+        "tokenizer.json",
+        "vocabulary.txt",
+    ]
+    if model_size == "large-v3" or "distil" in model_size:
+        files.append("preprocessor_config.json")
+        files.append("vocabulary.json")
+
+        files.remove("vocabulary.txt")
+
+    for attempt in range(2):
+        try:
+            snapshot_download(
+                repo_id=repo_id,
+                allow_patterns=files,
+                local_dir=model_path,
+            )
+            break
+        except LocalEntryNotFoundError:
+            if attempt < 1:
+                time.sleep(2)
+            else:
+                print("[ERROR] LocalEntryNotFoundError and no fallback.")
+                traceback.print_exc()
+                exit(1)
+        except Exception as e:
+            print(f"[ERROR] Unexpected error on attempt {attempt + 1}: {e}")
+            traceback.print_exc()
+            exit(1)
+
+    return model_path
+
+
+def execute_asr(input_folder, output_folder, model_path, language, precision):
+    if language == "auto":
+        language = None  # 不设置语种由模型自动输出概率最高的语种
+    print("loading faster whisper model:", model_path, model_path)
+    device = "cuda" if torch.cuda.is_available() else "cpu"
+    model = WhisperModel(model_path, device=device, compute_type=precision)
+
+    input_file_names = os.listdir(input_folder)
+    input_file_names.sort()
+
+    output = []
+    output_file_name = os.path.basename(input_folder)
+
+    for file_name in tqdm(input_file_names):
+        try:
+            file_path = os.path.join(input_folder, file_name)
+            segments, info = model.transcribe(
+                audio=file_path,
+                beam_size=5,
+                vad_filter=True,
+                vad_parameters=dict(min_silence_duration_ms=700),
+                language=language,
+            )
+            text = ""
+
+            if info.language == "zh":
+                print("检测为中文文本, 转 FunASR 处理")
+                text = only_asr(file_path, language=info.language.lower())
+
+            if text == "":
+                for segment in segments:
+                    text += segment.text
+            output.append(f"{file_path}|{output_file_name}|{info.language.upper()}|{text}")
+        except Exception as e:
+            print(e)
+            traceback.print_exc()
+
+    output_folder = output_folder or "output/asr_opt"
+    os.makedirs(output_folder, exist_ok=True)
+    output_file_path = os.path.abspath(f"{output_folder}/{output_file_name}.list")
+
+    with open(output_file_path, "w", encoding="utf-8") as f:
+        f.write("\n".join(output))
+        print(f"ASR 任务完成->标注文件路径: {output_file_path}\n")
+    return output_file_path
+
+
+load_cudnn()
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "-i", "--input_folder", type=str, required=True, help="Path to the folder containing WAV files."
+    )
+    parser.add_argument("-o", "--output_folder", type=str, required=True, help="Output folder to store transcriptions.")
+    parser.add_argument(
+        "-s",
+        "--model_size",
+        type=str,
+        default="large-v3",
+        choices=get_models(),
+        help="Model Size of Faster Whisper",
+    )
+    parser.add_argument(
+        "-l", "--language", type=str, default="ja", choices=language_code_list, help="Language of the audio files."
+    )
+    parser.add_argument(
+        "-p",
+        "--precision",
+        type=str,
+        default="float16",
+        choices=["float16", "float32", "int8"],
+        help="fp16, int8 or fp32",
+    )
+
+    cmd = parser.parse_args()
+    model_size = cmd.model_size
+    if model_size == "large":
+        model_size = "large-v3"
+    model_path = download_model(model_size)
+    output_file_path = execute_asr(
+        input_folder=cmd.input_folder,
+        output_folder=cmd.output_folder,
+        model_path=model_path,
+        language=cmd.language,
+        precision=cmd.precision,
+    )

ascend_910-gpt-sovits/GPT-SoVITS/tools/asr/funasr_asr.py

@@ -0,0 +1,118 @@
+# -*- coding:utf-8 -*-
+
+import argparse
+import os
+import traceback
+
+# from funasr.utils import version_checker
+# version_checker.check_for_update = lambda: None
+from funasr import AutoModel
+from tqdm import tqdm
+
+funasr_models = {}  # 存储模型避免重复加载
+
+
+def only_asr(input_file, language):
+    try:
+        model = create_model(language)
+        text = model.generate(input=input_file)[0]["text"]
+    except:
+        text = ""
+        print(traceback.format_exc())
+    return text
+
+
+def create_model(language="zh"):
+    path_vad = "tools/asr/models/speech_fsmn_vad_zh-cn-16k-common-pytorch"
+    path_punc = "tools/asr/models/punc_ct-transformer_zh-cn-common-vocab272727-pytorch"
+    path_vad = path_vad if os.path.exists(path_vad) else "iic/speech_fsmn_vad_zh-cn-16k-common-pytorch"
+    path_punc = path_punc if os.path.exists(path_punc) else "iic/punc_ct-transformer_zh-cn-common-vocab272727-pytorch"
+    vad_model_revision = punc_model_revision = "v2.0.4"
+
+    if language == "zh":
+        path_asr = "tools/asr/models/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch"
+        path_asr = (
+            path_asr
+            if os.path.exists(path_asr)
+            else "iic/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch"
+        )
+        model_revision = "v2.0.4"
+    elif language == "yue":
+        path_asr = "tools/asr/models/speech_UniASR_asr_2pass-cantonese-CHS-16k-common-vocab1468-tensorflow1-online"
+        path_asr = (
+            path_asr
+            if os.path.exists(path_asr)
+            else "iic/speech_UniASR_asr_2pass-cantonese-CHS-16k-common-vocab1468-tensorflow1-online"
+        )
+        model_revision = "master"
+        path_vad = path_punc = None
+        vad_model_revision = punc_model_revision = None
+        ###友情提示：粤语带VAD识别可能会有少量shape不对报错的，但是不带VAD可以.不带vad只能分阶段单独加标点。不过标点模型对粤语效果真的不行…
+    else:
+        raise ValueError("FunASR 不支持该语言" + ": " + language)
+
+    if language in funasr_models:
+        return funasr_models[language]
+    else:
+        model = AutoModel(
+            model=path_asr,
+            model_revision=model_revision,
+            vad_model=path_vad,
+            vad_model_revision=vad_model_revision,
+            punc_model=path_punc,
+            punc_model_revision=punc_model_revision,
+        )
+        print(f"FunASR 模型加载完成: {language.upper()}")
+
+        funasr_models[language] = model
+        return model
+
+
+def execute_asr(input_folder, output_folder, model_size, language):
+    input_file_names = os.listdir(input_folder)
+    input_file_names.sort()
+
+    output = []
+    output_file_name = os.path.basename(input_folder)
+
+    model = create_model(language)
+
+    for file_name in tqdm(input_file_names):
+        try:
+            print("\n" + file_name)
+            file_path = os.path.join(input_folder, file_name)
+            text = model.generate(input=file_path)[0]["text"]
+            output.append(f"{file_path}|{output_file_name}|{language.upper()}|{text}")
+        except:
+            print(traceback.format_exc())
+
+    output_folder = output_folder or "output/asr_opt"
+    os.makedirs(output_folder, exist_ok=True)
+    output_file_path = os.path.abspath(f"{output_folder}/{output_file_name}.list")
+
+    with open(output_file_path, "w", encoding="utf-8") as f:
+        f.write("\n".join(output))
+        print(f"ASR 任务完成->标注文件路径: {output_file_path}\n")
+    return output_file_path
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "-i", "--input_folder", type=str, required=True, help="Path to the folder containing WAV files."
+    )
+    parser.add_argument("-o", "--output_folder", type=str, required=True, help="Output folder to store transcriptions.")
+    parser.add_argument("-s", "--model_size", type=str, default="large", help="Model Size of FunASR is Large")
+    parser.add_argument(
+        "-l", "--language", type=str, default="zh", choices=["zh", "yue", "auto"], help="Language of the audio files."
+    )
+    parser.add_argument(
+        "-p", "--precision", type=str, default="float16", choices=["float16", "float32"], help="fp16 or fp32"
+    )  # 还没接入
+    cmd = parser.parse_args()
+    execute_asr(
+        input_folder=cmd.input_folder,
+        output_folder=cmd.output_folder,
+        model_size=cmd.model_size,
+        language=cmd.language,
+    )

ascend_910-gpt-sovits/GPT-SoVITS/tools/asr/models/.gitignore

@@ -0,0 +1,2 @@
+*
+!.gitignore