添加性能监控

Ref_Audio_Selector/common/time_util.py

@@ -0,0 +1,71 @@
+import time
+from Ref_Audio_Selector.config_param.log_config import p_logger
+import Ref_Audio_Selector.config_param.config_params as params
+
+
+def timeit_decorator(func):
+    """
+    装饰器，用于计算被装饰函数的执行时间。
+
+    参数:
+        func (function): 要计时的函数。
+
+    返回:
+        function: 包含计时功能的新函数。
+    """
+
+    def wrapper(*args, **kwargs):
+        if params.time_log_print_type != 'file':
+            return func(*args, **kwargs)
+
+        start_time = time.perf_counter()  # 使用 perf_counter 获取高精度计时起点
+
+        func_result = func(*args, **kwargs)  # 执行原函数
+
+        end_time = time.perf_counter()  # 获取计时终点
+        elapsed_time = end_time - start_time  # 计算执行耗时
+
+        # 记录日志内容
+        log_message = f"{func.__name__} 执行耗时: {elapsed_time:.6f} 秒"
+        p_logger.info(log_message)
+
+        return func_result
+
+    return wrapper
+
+
+def time_monitor(func):
+    """
+        返回结果，追加时间
+    """
+
+    def wrapper(*args, **kwargs):
+
+        start_time = time.perf_counter()  # 使用 perf_counter 获取高精度计时起点
+
+        func_result = func(*args, **kwargs)  # 执行原函数
+
+        end_time = time.perf_counter()  # 获取计时终点
+        elapsed_time = end_time - start_time  # 计算执行耗时
+
+        return elapsed_time, func_result
+
+    return wrapper
+
+
+# 使用装饰器
+@timeit_decorator
+def example_function(n):
+    time.sleep(n)  # 假设这是需要计时的函数，这里模拟耗时操作
+    return n * 2
+
+
+def example_function2(n):
+    time.sleep(n)  # 假设这是需要计时的函数，这里模拟耗时操作
+    return n * 2
+
+
+if __name__ == "__main__":
+    # 调用经过装饰的函数
+    # result = example_function(2)
+    print(time_monitor(example_function2)(2))

Ref_Audio_Selector/config.ini

@@ -1,6 +1,10 @@
 # config.ini
 
 [Base]
+# 函数时间消耗日志打印类型 file 打印到文件; close 关闭
+time_log_print_type = file
+# 函数时间消耗日志保存目录路径
+time_log_print_dir = Ref_Audio_Selector/log/performance
 # 参考音频目录
 reference_audio_dir = refer_audio
 

Ref_Audio_Selector/config_param/config_params.py

@@ -3,6 +3,10 @@ import Ref_Audio_Selector.config_param.config_manager as config_manager
 config = config_manager.get_config()
 
 # [Base]
+# 函数时间消耗日志打印类型 file 打印到文件; close 关闭
+time_log_print_type = config.get_base('time_log_print_type')
+# 函数时间消耗日志保存目录路径
+time_log_print_dir = config.get_base('time_log_print_dir')
 # 参考音频目录
 reference_audio_dir = config.get_base('reference_audio_dir')
 

Ref_Audio_Selector/config_param/log_config.py

@@ -0,0 +1,39 @@
+import logging
+import datetime
+import Ref_Audio_Selector.config_param.config_params as params
+
+
+def setup_logging():
+    # 获取当前日期，用于文件名和日志内容
+    current_date = datetime.datetime.now().strftime('%Y-%m-%d')
+
+    # 创建一个用于常规日志的处理器
+    general_handler = logging.FileHandler('general.log', mode='a', encoding='utf-8')
+    general_handler.setLevel(logging.INFO)
+    general_formatter = logging.Formatter('%(asctime)s - %(levelname)s - %(message)s')
+    general_handler.setFormatter(general_formatter)
+
+    # 创建一个专用于性能监控日志的处理器
+    performance_handler = logging.FileHandler(
+        f"{params.time_log_print_dir}/{current_date}.log", mode='a', encoding='utf-8')
+    performance_handler.setLevel(logging.INFO)
+    performance_formatter = logging.Formatter('%(asctime)s - %(levelname)s - %(message)s')
+    performance_handler.setFormatter(performance_formatter)
+
+    # 配置一个常规的logger
+    general_logger = logging.getLogger('general')
+    general_logger.setLevel(logging.INFO)
+    general_logger.addHandler(general_handler)
+
+    # 配置一个专门用于性能监控的logger
+    performance_logger = logging.getLogger('performance')
+    performance_logger.setLevel(logging.INFO)
+    performance_logger.addHandler(performance_handler)
+
+    # 配置根logger，以防万一
+    logging.basicConfig(level=logging.WARNING, handlers=[general_handler])
+
+    return general_logger, performance_logger
+
+
+logger, p_logger = setup_logging()

Ref_Audio_Selector/ref_audio_selector_webui.py

@@ -9,6 +9,8 @@ import Ref_Audio_Selector.tool.audio_config as audio_config
 import Ref_Audio_Selector.tool.delete_inference_with_ref as delete_inference_with_ref
 import Ref_Audio_Selector.common.common as common
 import Ref_Audio_Selector.config_param.config_params as params
+import Ref_Audio_Selector.common.time_util as time_util
+from Ref_Audio_Selector.config_param.log_config import logger
 from tools.i18n.i18n import I18nAuto
 from config import python_exec, is_half
 from tools import my_utils
@@ -16,11 +18,11 @@ from tools.asr.config import asr_dict
 from subprocess import Popen
 
 i18n = I18nAuto()
+rw_param = params.config_manager.get_rw_param()
 
 p_similarity = None
 p_asr = None
 p_text_similarity = None
-rw_param = params.config_manager.get_rw_param()
 
 
 # 校验基础信息
@@ -50,10 +52,13 @@ def convert_from_list(text_work_space_dir, text_role, text_list_input):
 
         ref_audio_all = os.path.join(base_role_dir,
                                      params.list_to_convert_reference_audio_dir)
-        text_convert_from_list_info = f"转换成功：生成目录{ref_audio_all}"
+
+        time_consuming, _ = time_util.time_monitor(audio_similarity.convert_from_list)(text_list_input, ref_audio_all)
+
+        text_convert_from_list_info = f"耗时：{time_consuming:0.1f}秒；转换成功：生成目录{ref_audio_all}"
         text_sample_dir = ref_audio_all
 
-        audio_similarity.convert_from_list(text_list_input, ref_audio_all)
+        # audio_similarity.convert_from_list(text_list_input, ref_audio_all)
     except Exception as e:
         traceback.print_exc()
         text_convert_from_list_info = f"发生异常：{e}"
@@ -113,10 +118,17 @@ def sample(text_work_space_dir, text_role, text_sample_dir, text_base_voice_path
             raise Exception("每段随机抽样个数不能为空")
 
         ref_audio_dir = os.path.join(base_role_dir, params.reference_audio_dir)
-        text_sample_info = f"抽样成功：生成目录{ref_audio_dir}"
 
-        similarity_list, _, _ = start_similarity_analysis(base_role_dir, text_sample_dir,
+        time_consuming, similarity_list, _, _ = (
-                                                          text_base_voice_path, checkbox_similarity_output)
+            time_util.time_monitor(start_similarity_analysis)(base_role_dir,
+                                                              text_sample_dir,
+                                                              text_base_voice_path,
+                                                              checkbox_similarity_output))
+
+        text_sample_info = f"耗时：{time_consuming:0.1f}秒；抽样成功：生成目录{ref_audio_dir}"
+
+        # similarity_list, _, _ = start_similarity_analysis(base_role_dir, text_sample_dir,
+        #                                                   text_base_voice_path, checkbox_similarity_output)
 
         if similarity_list is None:
             raise Exception("相似度分析失败")
@@ -136,9 +148,9 @@ def sample(text_work_space_dir, text_role, text_sample_dir, text_base_voice_path
 # 根据参考音频和测试文本，执行批量推理
 def model_inference(text_work_space_dir, text_role, text_model_inference_voice_dir, text_url,
                     text_text, text_ref_path, text_ref_text, text_emotion,
-                    text_test_content):
+                    text_test_content_dir):
-    text_work_space_dir, text_model_inference_voice_dir, text_test_content \
+    text_work_space_dir, text_model_inference_voice_dir, text_test_content_dir \
-        = common.batch_clean_paths([text_work_space_dir, text_model_inference_voice_dir, text_test_content])
+        = common.batch_clean_paths([text_work_space_dir, text_model_inference_voice_dir, text_test_content_dir])
 
     inference_dir = None
     text_asr_audio_dir = None
@@ -151,7 +163,7 @@ def model_inference(text_work_space_dir, text_role, text_model_inference_voice_d
             raise Exception("推理服务请求地址不能为空")
         if text_text is None or text_text == '':
             raise Exception("文本参数名不能为空")
-        if text_test_content is None or text_test_content == '':
+        if text_test_content_dir is None or text_test_content_dir == '':
             raise Exception("待推理文本路径不能为空")
         if (text_ref_path is None or text_ref_path == '') and (text_ref_text is None or text_ref_text == '') and (
                 text_emotion is None or text_emotion == ''):
@@ -160,18 +172,24 @@ def model_inference(text_work_space_dir, text_role, text_model_inference_voice_d
         inference_dir = os.path.join(base_role_dir, params.inference_audio_dir)
         text_asr_audio_dir = os.path.join(inference_dir,
                                           params.inference_audio_text_aggregation_dir)
-        text_model_inference_info = f"推理成功：生成目录{inference_dir}"
 
         url_composer = audio_inference.URLComposer(text_url, text_emotion, text_text, text_ref_path, text_ref_text)
         url_composer.is_valid()
-        text_list = common.read_text_file_to_list(text_test_content)
+        text_list = common.read_text_file_to_list(text_test_content_dir)
         if text_list is None or len(text_list) == 0:
             raise Exception("待推理文本内容不能为空")
         ref_audio_manager = common.RefAudioListManager(text_model_inference_voice_dir)
         if len(ref_audio_manager.get_audio_list()) == 0:
             raise Exception("待推理的参考音频不能为空")
-        audio_inference.generate_audio_files(url_composer, text_list, ref_audio_manager.get_ref_audio_list(),
+
-                                             inference_dir)
+        time_consuming, _ = time_util.time_monitor(audio_inference.generate_audio_files)(url_composer, text_list,
+                                                                                         ref_audio_manager.get_ref_audio_list(),
+                                                                                         inference_dir)
+
+        text_model_inference_info = f"耗时：{time_consuming:0.1f}秒；推理成功：生成目录{inference_dir}"
+
+        # audio_inference.generate_audio_files(url_composer, text_list, ref_audio_manager.get_ref_audio_list(),
+        #                                      inference_dir)
     except Exception as e:
         traceback.print_exc()
         text_model_inference_info = f"发生异常：{e}"
@@ -198,10 +216,15 @@ def asr(text_work_space_dir, text_role, text_asr_audio_dir, dropdown_asr_model,
             raise Exception("asr模型大小不能为空")
         if dropdown_asr_lang is None or dropdown_asr_lang == '':
             raise Exception("asr语言不能为空")
-        asr_file = open_asr(text_asr_audio_dir, base_role_dir, dropdown_asr_model, dropdown_asr_size,
+
-                            dropdown_asr_lang)
+        time_consuming, asr_file = time_util.time_monitor(open_asr)(text_asr_audio_dir, base_role_dir,
+                                                                    dropdown_asr_model, dropdown_asr_size,
+                                                                    dropdown_asr_lang)
+
+        # asr_file = open_asr(text_asr_audio_dir, base_role_dir, dropdown_asr_model, dropdown_asr_size,
+        #                     dropdown_asr_lang)
         text_text_similarity_analysis_path = asr_file
-        text_asr_info = f"asr成功：生成文件{asr_file}"
+        text_asr_info = f"耗时：{time_consuming:0.1f}秒；asr成功：生成文件{asr_file}"
     except Exception as e:
         traceback.print_exc()
         text_asr_info = f"发生异常：{e}"
@@ -253,8 +276,13 @@ def text_similarity_analysis(text_work_space_dir, text_role,
         if text_text_similarity_analysis_path is None or text_text_similarity_analysis_path == '':
             raise Exception("asr生成的文件路径不能为空，请先完成上一步操作")
         similarity_dir = os.path.join(base_role_dir, params.text_similarity_output_dir)
-        text_text_similarity_analysis_info = f"相似度分析成功：生成目录{similarity_dir}"
+
-        open_text_similarity_analysis(text_text_similarity_analysis_path, similarity_dir)
+        time_consuming, _ = time_util.time_monitor(open_text_similarity_analysis)(text_text_similarity_analysis_path,
+                                                                                  similarity_dir)
+
+        text_text_similarity_analysis_info = f"耗时：{time_consuming:0.1f}秒；相似度分析成功：生成目录{similarity_dir}"
+
+        # open_text_similarity_analysis(text_text_similarity_analysis_path, similarity_dir)
     except Exception as e:
         traceback.print_exc()
         text_text_similarity_analysis_info = f"发生异常：{e}"
@@ -293,13 +321,18 @@ def similarity_audio_output(text_work_space_dir, text_role, text_base_audio_path
             raise Exception("基准音频路径不能为空")
         if text_compare_audio_dir is None or text_compare_audio_dir == '':
             raise Exception("待分析的音频所在目录不能为空")
-        similarity_list, similarity_file, similarity_file_dir = start_similarity_analysis(
+
-            base_role_dir, text_compare_audio_dir, text_base_audio_path, True)
+        time_consuming, similarity_list, similarity_file, similarity_file_dir \
+            = time_util.time_monitor(start_similarity_analysis)(base_role_dir,
+                                                                text_compare_audio_dir, text_base_audio_path, True)
+
+        # similarity_list, similarity_file, similarity_file_dir = start_similarity_analysis(
+        #     base_role_dir, text_compare_audio_dir, text_base_audio_path, True)
 
         if similarity_list is None:
             raise Exception("相似度分析失败")
 
-        text_similarity_audio_output_info = f'相似度分析成功：生成目录{similarity_file_dir}，文件{similarity_file}'
+        text_similarity_audio_output_info = f'耗时：{time_consuming:0.1f}秒；相似度分析成功：生成目录{similarity_file_dir}，文件{similarity_file}'
 
     except Exception as e:
         traceback.print_exc()
@@ -320,9 +353,13 @@ def sync_ref_audio(text_work_space_dir, text_role, text_sync_ref_audio_dir,
             raise Exception("参考音频目录不能为空")
         if text_sync_inference_audio_dir is None or text_sync_inference_audio_dir == '':
             raise Exception("推理生成的音频目录不能为空")
-        delete_text_wav_num, delete_emotion_dir_num = delete_inference_with_ref.sync_ref_audio(text_sync_ref_audio_dir,
+        time_consuming, delete_text_wav_num, delete_emotion_dir_num \
-                                                                                               text_sync_inference_audio_dir)
+            = time_util.time_monitor(delete_inference_with_ref.sync_ref_audio)(text_sync_ref_audio_dir,
-        text_sync_ref_audio_info = f"推理音频目录{text_sync_inference_audio_dir}下，text目录删除了{delete_text_wav_num}个参考音频，emotion目录下，删除了{delete_emotion_dir_num}个目录"
+                                                                               text_sync_inference_audio_dir)
+        # delete_text_wav_num, delete_emotion_dir_num = delete_inference_with_ref.sync_ref_audio(
+        # text_sync_ref_audio_dir, text_sync_inference_audio_dir)
+        text_sync_ref_audio_info = (f"耗时：{time_consuming:0.1f}秒；推理音频目录{text_sync_inference_audio_dir}下，"
+                                    f"text目录删除了{delete_text_wav_num}个参考音频，emotion目录下，删除了{delete_emotion_dir_num}个目录")
     except Exception as e:
         traceback.print_exc()
         text_sync_ref_audio_info = f"发生异常：{e}"
@@ -343,10 +380,17 @@ def create_config(text_work_space_dir, text_role, text_template, text_sync_ref_a
         if text_sync_ref_audio_dir2 is None or text_sync_ref_audio_dir2 == '':
             raise Exception("参考音频目录不能为空")
         config_file = os.path.join(base_role_dir, f'{params.reference_audio_config_filename}.json')
-        text_create_config_info = f"配置生成成功：生成文件{config_file}"
         ref_audio_manager = common.RefAudioListManager(text_sync_ref_audio_dir2)
-        audio_config.generate_audio_config(base_role_dir, text_template, ref_audio_manager.get_ref_audio_list(),
+
-                                           config_file)
+        time_consuming, _ = time_util.time_monitor(audio_config.generate_audio_config)(base_role_dir, text_template,
+                                                                                       ref_audio_manager.get_ref_audio_list(),
+                                                                                       config_file)
+
+        # audio_config.generate_audio_config(base_role_dir, text_template, ref_audio_manager.get_ref_audio_list(),
+        #                                    config_file)
+
+        text_create_config_info = f"耗时：{time_consuming:0.1f}秒；配置生成成功：生成文件{config_file}"
+
     except Exception as e:
         traceback.print_exc()
         text_create_config_info = f"发生异常：{e}"
@@ -514,7 +558,8 @@ with gr.Blocks() as app:
     button_model_inference.click(model_inference,
                                  [text_work_space_dir, text_role, text_model_inference_voice_dir, text_url,
                                   text_text, text_ref_path, text_ref_text, text_emotion,
-                                  text_test_content], [text_model_inference_info, text_asr_audio_dir, text_sync_inference_audio_dir])
+                                  text_test_content],
+                                 [text_model_inference_info, text_asr_audio_dir, text_sync_inference_audio_dir])
 
 app.launch(
     server_port=9423,

Ref_Audio_Selector/tool/asr/funasr_asr_multi_level_dir.py

@@ -4,6 +4,7 @@ import argparse
 import os
 import traceback
 import Ref_Audio_Selector.config_param.config_params as params
+from Ref_Audio_Selector.common.time_util import timeit_decorator
 from tqdm import tqdm
 from funasr import AutoModel
 
@@ -34,30 +35,7 @@ def only_asr(input_file):
     return text
 
 
-def execute_asr(input_folder, output_folder, model_size, language):
+@timeit_decorator
-    input_file_names = os.listdir(input_folder)
-    input_file_names.sort()
-
-    output = []
-    output_file_name = os.path.basename(input_folder)
-
-    for name in tqdm(input_file_names):
-        try:
-            text = model.generate(input="%s/%s" % (input_folder, name))[0]["text"]
-            output.append(f"{input_folder}/{name}|{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
-
-
 def execute_asr_multi_level_dir(input_folder, output_folder, model_size, language):
     output = []
     output_file_name = os.path.basename(input_folder)

Ref_Audio_Selector/tool/audio_inference.py

@@ -2,6 +2,7 @@ import os
 import requests
 import itertools
 import Ref_Audio_Selector.config_param.config_params as params
+from Ref_Audio_Selector.common.time_util import timeit_decorator
 from urllib.parse import urlparse, parse_qs, urlencode, urlunparse, quote
 
 
@@ -72,6 +73,7 @@ def safe_encode_query_params(original_url):
     return encoded_url
 
 
+@timeit_decorator
 def generate_audio_files(url_composer, text_list, emotion_list, output_dir_path):
     # Ensure the output directory exists
     output_dir = os.path.abspath(output_dir_path)

Ref_Audio_Selector/tool/audio_similarity.py

@@ -6,11 +6,9 @@ import librosa
 
 def check_audio_duration(path, min_duration=3, max_duration=10):
     try:
-        # 加载音频文件
-        audio, sample_rate = librosa.load(path)
 
-        # 计算音频的时长（单位：秒）
+        # 直接计算音频文件的时长（单位：秒）
-        duration = librosa.get_duration(y=audio, sr=sample_rate)
+        duration = librosa.get_duration(filename=path)
 
         # 判断时长是否在3s至10s之间
         if min_duration <= duration <= max_duration:

Ref_Audio_Selector/tool/speaker_verification/voice_similarity.py

@@ -1,5 +1,6 @@
 import argparse
 import os
+from Ref_Audio_Selector.common.time_util import timeit_decorator
 
 from modelscope.pipelines import pipeline
 
@@ -10,6 +11,7 @@ sv_pipeline = pipeline(
 )
 
 
+@timeit_decorator
 def compare_audio_and_generate_report(reference_audio_path, comparison_dir_path, output_file_path):
     # Step 1: 获取比较音频目录下所有音频文件的路径
     comparison_audio_paths = [os.path.join(comparison_dir_path, f) for f in os.listdir(comparison_dir_path) if

Ref_Audio_Selector/tool/text_comparison/asr_text_process.py

@@ -2,6 +2,7 @@ import os
 import argparse
 from collections import defaultdict
 from operator import itemgetter
+from Ref_Audio_Selector.common.time_util import timeit_decorator
 import Ref_Audio_Selector.tool.text_comparison.text_comparison as text_comparison
 import Ref_Audio_Selector.config_param.config_params as params
 import Ref_Audio_Selector.common.common as common
@@ -88,6 +89,7 @@ def format_list_to_text(data_list, output_filename):
                 output_file.write(formatted_line)
 
 
+@timeit_decorator
 def process(asr_file_path, output_dir, similarity_enlarge_boundary):
     # 检查输出目录是否存在，如果不存在则创建
     if not os.path.exists(output_dir):