mirror of
https://github.com/RVC-Boss/GPT-SoVITS.git
synced 2025-04-05 19:41:56 +08:00
api推理，添加多进程请求
This commit is contained in:
Downupanddownup
parent
2a23f95f61
commit
c36d0a93fe
@ -1,4 +1,5 @@
|
||||
import time
|
||||
import os
|
||||
from Ref_Audio_Selector.config_param.log_config import p_logger
|
||||
import Ref_Audio_Selector.config_param.config_params as params
|
||||
|
||||
@ -26,7 +27,7 @@ def timeit_decorator(func):
|
||||
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"
|
||||
log_message = f"进程ID: {os.getpid()}, {func.__name__} 执行耗时: {elapsed_time:.6f} 秒"
|
||||
p_logger.info(log_message)
|
||||
|
||||
return func_result
|
||||
|
||||
@ -13,6 +13,7 @@ def create_general_logger():
|
||||
# 可以设置控制台输出的格式
|
||||
console_formatter = logging.Formatter('%(asctime)s - %(levelname)s - %(message)s')
|
||||
console_handler.setFormatter(console_formatter)
|
||||
console_handler.encoding = 'utf-8' # 设置字符编码为utf-8
|
||||
|
||||
# 创建一个用于常规日志的处理器
|
||||
general_handler = logging.FileHandler(f"{params.log_dir}/{current_date}.log", mode='a', encoding='utf-8')
|
||||
|
||||
@ -184,9 +184,10 @@ def model_inference(text_work_space_dir, text_role, text_model_inference_voice_d
|
||||
if len(ref_audio_manager.get_audio_list()) == 0:
|
||||
raise Exception("待推理的参考音频不能为空")
|
||||
|
||||
time_consuming, _ = time_util.time_monitor(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_parallel)(url_composer,
|
||||
text_list,
|
||||
ref_audio_manager.get_ref_audio_list(),
|
||||
inference_dir, 3)
|
||||
|
||||
text_model_inference_info = f"耗时：{time_consuming:0.1f}秒；推理成功：生成目录{inference_dir}"
|
||||
|
||||
@ -449,171 +450,176 @@ def save_role(text_role):
|
||||
rw_param.write(rw_param.role, text_role)
|
||||
|
||||
|
||||
default_work_space_dir = rw_param.read(rw_param.work_dir)
|
||||
default_role = rw_param.read(rw_param.role)
|
||||
default_base_dir = os.path.join(default_work_space_dir, default_role)
|
||||
if __name__ == '__main__':
|
||||
default_work_space_dir = rw_param.read(rw_param.work_dir)
|
||||
default_role = rw_param.read(rw_param.role)
|
||||
default_base_dir = os.path.join(default_work_space_dir, default_role)
|
||||
|
||||
with gr.Blocks() as app:
|
||||
gr.Markdown(value=i18n("基本介绍：这是一个从训练素材中，批量提取参考音频，并进行效果评估与配置生成的工具"))
|
||||
with gr.Row():
|
||||
text_work_space_dir = gr.Text(label=i18n("工作目录，后续操作所生成文件都会保存在此目录下"),
|
||||
value=default_work_space_dir)
|
||||
text_role = gr.Text(label=i18n("角色名称"), value=default_role)
|
||||
text_work_space_dir.input(save_work_dir, [text_work_space_dir, text_role], [text_role])
|
||||
text_role.input(save_role, [text_role], [])
|
||||
with gr.Tab(label=i18n("第一步：基于训练素材，生成待选参考音频列表"), open=False):
|
||||
gr.Markdown(value=i18n("1.1：选择list文件，并提取3-10秒的素材作为参考候选"))
|
||||
text_list_input = gr.Text(label=i18n("请输入list文件路径"), value="")
|
||||
with gr.Blocks() as app:
|
||||
gr.Markdown(value=i18n("基本介绍：这是一个从训练素材中，批量提取参考音频，并进行效果评估与配置生成的工具"))
|
||||
with gr.Row():
|
||||
button_convert_from_list = gr.Button(i18n("开始生成待参考列表"), variant="primary")
|
||||
text_convert_from_list_info = gr.Text(label=i18n("参考列表生成结果"), value="", interactive=False)
|
||||
gr.Markdown(value=i18n("1.2：选择基准音频，执行相似度匹配，并分段随机抽样"))
|
||||
default_sample_dir = common.check_path_existence_and_return(
|
||||
os.path.join(default_base_dir, params.list_to_convert_reference_audio_dir))
|
||||
text_sample_dir = gr.Text(label=i18n("参考音频抽样目录"), value=default_sample_dir, interactive=True)
|
||||
button_convert_from_list.click(convert_from_list, [text_work_space_dir, text_role, text_list_input],
|
||||
[text_convert_from_list_info, text_sample_dir])
|
||||
with gr.Row():
|
||||
text_base_voice_path = gr.Text(label=i18n("请输入基准音频路径"), value="")
|
||||
text_subsection_num = gr.Text(label=i18n("请输入分段数"), value="10")
|
||||
text_sample_num = gr.Text(label=i18n("请输入每段随机抽样个数"), value="4")
|
||||
checkbox_similarity_output = gr.Checkbox(label=i18n("是否将相似度匹配结果输出到临时目录？"), show_label=True)
|
||||
with gr.Row():
|
||||
button_sample = gr.Button(i18n("开始分段随机抽样"), variant="primary")
|
||||
text_sample_info = gr.Text(label=i18n("分段随机抽样结果"), value="", interactive=False)
|
||||
with gr.Tab(label=i18n("第二步：基于参考音频和测试文本，执行批量推理"), open=False):
|
||||
gr.Markdown(value=i18n("2.1：配置推理服务参数信息，参考音频路径/文本和角色情绪二选一，如果是角色情绪，需要先执行第四步，"
|
||||
"将参考音频打包配置到推理服务下，在推理前，请确认完整请求地址是否与正常使用时的一致，包括角色名称，尤其是文本分隔符是否正确"))
|
||||
default_model_inference_voice_dir = common.check_path_existence_and_return(
|
||||
os.path.join(default_base_dir, params.reference_audio_dir))
|
||||
text_model_inference_voice_dir = gr.Text(label=i18n("待推理的参考音频所在目录"),
|
||||
value=default_model_inference_voice_dir, interactive=True)
|
||||
text_url = gr.Text(label=i18n("请输入推理服务请求地址与参数"), value=rw_param.read(rw_param.generate_audio_url))
|
||||
with gr.Row():
|
||||
text_text = gr.Text(label=i18n("请输入文本参数名"), value=rw_param.read(rw_param.text_param))
|
||||
text_ref_path = gr.Text(label=i18n("请输入参考音频路径参数名"),
|
||||
value=rw_param.read(rw_param.ref_path_param))
|
||||
text_ref_text = gr.Text(label=i18n("请输入参考音频文本参数名"),
|
||||
value=rw_param.read(rw_param.ref_text_param))
|
||||
text_emotion = gr.Text(label=i18n("请输入角色情绪参数名"), value=rw_param.read(rw_param.emotion_param))
|
||||
text_whole_url = gr.Text(label=i18n("完整地址"), value="", interactive=False)
|
||||
text_url.input(whole_url, [text_url, text_text, text_ref_path, text_ref_text, text_emotion],
|
||||
[text_whole_url])
|
||||
text_url.blur(save_generate_audio_url, [text_url], [])
|
||||
text_text.input(whole_url, [text_url, text_text, text_ref_path, text_ref_text, text_emotion],
|
||||
[text_whole_url])
|
||||
text_text.blur(save_text_param, [text_text], [])
|
||||
text_ref_path.input(whole_url, [text_url, text_text, text_ref_path, text_ref_text, text_emotion],
|
||||
[text_whole_url])
|
||||
text_ref_path.blur(save_ref_path_param, [text_ref_path], [])
|
||||
text_ref_text.input(whole_url, [text_url, text_text, text_ref_path, text_ref_text, text_emotion],
|
||||
[text_whole_url])
|
||||
text_ref_text.blur(save_ref_text_param, [text_ref_text], [])
|
||||
text_emotion.input(whole_url, [text_url, text_text, text_ref_path, text_ref_text, text_emotion],
|
||||
[text_whole_url])
|
||||
text_emotion.blur(save_emotion_param, [text_emotion], [])
|
||||
gr.Markdown(value=i18n("2.2：配置待推理文本，一句一行，不要太多，10条即可"))
|
||||
default_test_content_path = params.default_test_text_path
|
||||
text_test_content = gr.Text(label=i18n("请输入待推理文本路径"), value=default_test_content_path)
|
||||
gr.Markdown(value=i18n("2.3：启动推理服务，如果还没启动的话"))
|
||||
gr.Markdown(value=i18n("2.4：开始批量推理，这个过程比较耗时，可以去干点别的"))
|
||||
with gr.Row():
|
||||
button_model_inference = gr.Button(i18n("开启批量推理"), variant="primary")
|
||||
text_model_inference_info = gr.Text(label=i18n("批量推理结果"), value="", interactive=False)
|
||||
with gr.Tab(label=i18n("第三步：进行参考音频效果校验与筛选"), open=False):
|
||||
gr.Markdown(value=i18n("3.1：启动asr，获取推理音频文本"))
|
||||
default_asr_audio_dir = common.check_path_existence_and_return(
|
||||
os.path.join(default_base_dir, params.inference_audio_dir, params.inference_audio_text_aggregation_dir))
|
||||
text_asr_audio_dir = gr.Text(label=i18n("待asr的音频所在目录"), value=default_asr_audio_dir, interactive=True)
|
||||
with gr.Row():
|
||||
dropdown_asr_model = gr.Dropdown(
|
||||
label=i18n("ASR 模型"),
|
||||
choices=[],
|
||||
interactive=True,
|
||||
value="达摩 ASR (中文)"
|
||||
)
|
||||
dropdown_asr_size = gr.Dropdown(
|
||||
label=i18n("ASR 模型尺寸"),
|
||||
choices=["large"],
|
||||
interactive=True,
|
||||
value="large"
|
||||
)
|
||||
dropdown_asr_lang = gr.Dropdown(
|
||||
label=i18n("ASR 语言设置"),
|
||||
choices=["zh"],
|
||||
interactive=True,
|
||||
value="zh"
|
||||
)
|
||||
with gr.Row():
|
||||
button_asr = gr.Button(i18n("启动asr"), variant="primary")
|
||||
text_asr_info = gr.Text(label=i18n("asr结果"), value="", interactive=False)
|
||||
gr.Markdown(value=i18n("3.2：启动文本相似度分析"))
|
||||
default_text_similarity_analysis_path = common.check_path_existence_and_return(
|
||||
os.path.join(default_base_dir, params.asr_filename + '.list'))
|
||||
text_text_similarity_analysis_path = gr.Text(label=i18n("待分析的文件路径"),
|
||||
value=default_text_similarity_analysis_path, interactive=True)
|
||||
button_asr.click(asr, [text_work_space_dir, text_role, text_asr_audio_dir, dropdown_asr_model,
|
||||
dropdown_asr_size, dropdown_asr_lang],
|
||||
[text_asr_info, text_text_similarity_analysis_path])
|
||||
with gr.Row():
|
||||
button_text_similarity_analysis = gr.Button(i18n("启动文本相似度分析"), variant="primary")
|
||||
text_text_similarity_analysis_info = gr.Text(label=i18n("文本相似度分析结果"), value="", interactive=False)
|
||||
button_text_similarity_analysis.click(text_similarity_analysis, [text_work_space_dir, text_role,
|
||||
text_text_similarity_analysis_path],
|
||||
[text_text_similarity_analysis_info])
|
||||
gr.Markdown(value=i18n("3.3：根据相似度分析结果，重点检查最后几条是否存在复读等问题"))
|
||||
gr.Markdown(value=i18n("3.4：对结果按音频相似度排序，筛选低音质音频"))
|
||||
with gr.Row():
|
||||
text_base_audio_path = gr.Text(label=i18n("请输入基准音频"), value="")
|
||||
text_compare_audio_dir = gr.Text(label=i18n("请输入待比较的音频文件目录"), value="")
|
||||
with gr.Row():
|
||||
button_similarity_audio_output = gr.Button(i18n("输出相似度-参考音频到临时目录"), variant="primary")
|
||||
text_similarity_audio_output_info = gr.Text(label=i18n("输出结果"), value="", interactive=False)
|
||||
button_similarity_audio_output.click(similarity_audio_output,
|
||||
[text_work_space_dir, text_role, text_base_audio_path,
|
||||
text_compare_audio_dir], [text_similarity_audio_output_info])
|
||||
with gr.Row():
|
||||
default_sync_ref_audio_dir = common.check_path_existence_and_return(
|
||||
text_work_space_dir = gr.Text(label=i18n("工作目录，后续操作所生成文件都会保存在此目录下"),
|
||||
value=default_work_space_dir)
|
||||
text_role = gr.Text(label=i18n("角色名称"), value=default_role)
|
||||
text_work_space_dir.input(save_work_dir, [text_work_space_dir, text_role], [text_role])
|
||||
text_role.input(save_role, [text_role], [])
|
||||
with gr.Tab(label=i18n("第一步：基于训练素材，生成待选参考音频列表"), open=False):
|
||||
gr.Markdown(value=i18n("1.1：选择list文件，并提取3-10秒的素材作为参考候选"))
|
||||
text_list_input = gr.Text(label=i18n("请输入list文件路径"), value="")
|
||||
with gr.Row():
|
||||
button_convert_from_list = gr.Button(i18n("开始生成待参考列表"), variant="primary")
|
||||
text_convert_from_list_info = gr.Text(label=i18n("参考列表生成结果"), value="", interactive=False)
|
||||
gr.Markdown(value=i18n("1.2：选择基准音频，执行相似度匹配，并分段随机抽样"))
|
||||
default_sample_dir = common.check_path_existence_and_return(
|
||||
os.path.join(default_base_dir, params.list_to_convert_reference_audio_dir))
|
||||
text_sample_dir = gr.Text(label=i18n("参考音频抽样目录"), value=default_sample_dir, interactive=True)
|
||||
button_convert_from_list.click(convert_from_list, [text_work_space_dir, text_role, text_list_input],
|
||||
[text_convert_from_list_info, text_sample_dir])
|
||||
with gr.Row():
|
||||
text_base_voice_path = gr.Text(label=i18n("请输入基准音频路径"), value="")
|
||||
text_subsection_num = gr.Text(label=i18n("请输入分段数"), value="10")
|
||||
text_sample_num = gr.Text(label=i18n("请输入每段随机抽样个数"), value="4")
|
||||
checkbox_similarity_output = gr.Checkbox(label=i18n("是否将相似度匹配结果输出到临时目录？"),
|
||||
show_label=True)
|
||||
with gr.Row():
|
||||
button_sample = gr.Button(i18n("开始分段随机抽样"), variant="primary")
|
||||
text_sample_info = gr.Text(label=i18n("分段随机抽样结果"), value="", interactive=False)
|
||||
with gr.Tab(label=i18n("第二步：基于参考音频和测试文本，执行批量推理"), open=False):
|
||||
gr.Markdown(value=i18n("2.1：配置推理服务参数信息，参考音频路径/文本和角色情绪二选一，如果是角色情绪，需要先执行第四步，"
|
||||
"将参考音频打包配置到推理服务下，在推理前，请确认完整请求地址是否与正常使用时的一致，包括角色名称，尤其是文本分隔符是否正确"))
|
||||
default_model_inference_voice_dir = common.check_path_existence_and_return(
|
||||
os.path.join(default_base_dir, params.reference_audio_dir))
|
||||
text_sync_ref_audio_dir = gr.Text(label=i18n("参考音频路径"), value=default_sync_ref_audio_dir,
|
||||
interactive=True)
|
||||
default_sync_inference_audio_dir = common.check_path_existence_and_return(
|
||||
os.path.join(default_base_dir, params.inference_audio_dir))
|
||||
text_sync_inference_audio_dir = gr.Text(label=i18n("被同步的推理音频路径"),
|
||||
value=default_sync_inference_audio_dir, interactive=True)
|
||||
with gr.Row():
|
||||
button_sync_ref_audio = gr.Button(i18n("将参考音频的删除情况，同步到推理音频目录"), variant="primary")
|
||||
text_sync_ref_info = gr.Text(label=i18n("同步结果"), value="", interactive=False)
|
||||
button_sync_ref_audio.click(sync_ref_audio, [text_work_space_dir, text_role, text_sync_ref_audio_dir,
|
||||
text_sync_inference_audio_dir], [text_sync_ref_info])
|
||||
with gr.Tab("第四步：生成参考音频配置文本", open=False):
|
||||
gr.Markdown(value=i18n("4.1：编辑模板"))
|
||||
default_template_path = params.default_template_path
|
||||
default_template_content = common.read_file(default_template_path)
|
||||
text_template_path = gr.Text(label=i18n("模板文件路径"), value=default_template_path, interactive=True)
|
||||
text_template = gr.Text(label=i18n("模板内容"), value=default_template_content, lines=10)
|
||||
gr.Markdown(value=i18n("4.2：生成配置"))
|
||||
default_sync_ref_audio_dir2 = common.check_path_existence_and_return(
|
||||
os.path.join(default_base_dir, params.reference_audio_dir))
|
||||
text_sync_ref_audio_dir2 = gr.Text(label=i18n("参考音频路径"), value=default_sync_ref_audio_dir2,
|
||||
interactive=True)
|
||||
with gr.Row():
|
||||
button_create_config = gr.Button(i18n("生成配置"), variant="primary")
|
||||
text_create_config_info = gr.Text(label=i18n("生成结果"), value="", interactive=False)
|
||||
button_create_config.click(create_config,
|
||||
[text_work_space_dir, text_role, text_template, text_sync_ref_audio_dir2],
|
||||
[text_create_config_info])
|
||||
button_sample.click(sample, [text_work_space_dir, text_role, text_sample_dir, text_base_voice_path,
|
||||
text_subsection_num, text_sample_num, checkbox_similarity_output],
|
||||
[text_sample_info, text_model_inference_voice_dir, text_sync_ref_audio_dir,
|
||||
text_sync_ref_audio_dir2])
|
||||
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_model_inference_voice_dir = gr.Text(label=i18n("待推理的参考音频所在目录"),
|
||||
value=default_model_inference_voice_dir, interactive=True)
|
||||
text_url = gr.Text(label=i18n("请输入推理服务请求地址与参数"),
|
||||
value=rw_param.read(rw_param.generate_audio_url))
|
||||
with gr.Row():
|
||||
text_text = gr.Text(label=i18n("请输入文本参数名"), value=rw_param.read(rw_param.text_param))
|
||||
text_ref_path = gr.Text(label=i18n("请输入参考音频路径参数名"),
|
||||
value=rw_param.read(rw_param.ref_path_param))
|
||||
text_ref_text = gr.Text(label=i18n("请输入参考音频文本参数名"),
|
||||
value=rw_param.read(rw_param.ref_text_param))
|
||||
text_emotion = gr.Text(label=i18n("请输入角色情绪参数名"), value=rw_param.read(rw_param.emotion_param))
|
||||
text_whole_url = gr.Text(label=i18n("完整地址"), value="", interactive=False)
|
||||
text_url.input(whole_url, [text_url, text_text, text_ref_path, text_ref_text, text_emotion],
|
||||
[text_whole_url])
|
||||
text_url.blur(save_generate_audio_url, [text_url], [])
|
||||
text_text.input(whole_url, [text_url, text_text, text_ref_path, text_ref_text, text_emotion],
|
||||
[text_whole_url])
|
||||
text_text.blur(save_text_param, [text_text], [])
|
||||
text_ref_path.input(whole_url, [text_url, text_text, text_ref_path, text_ref_text, text_emotion],
|
||||
[text_whole_url])
|
||||
text_ref_path.blur(save_ref_path_param, [text_ref_path], [])
|
||||
text_ref_text.input(whole_url, [text_url, text_text, text_ref_path, text_ref_text, text_emotion],
|
||||
[text_whole_url])
|
||||
text_ref_text.blur(save_ref_text_param, [text_ref_text], [])
|
||||
text_emotion.input(whole_url, [text_url, text_text, text_ref_path, text_ref_text, text_emotion],
|
||||
[text_whole_url])
|
||||
text_emotion.blur(save_emotion_param, [text_emotion], [])
|
||||
gr.Markdown(value=i18n("2.2：配置待推理文本，一句一行，不要太多，10条即可"))
|
||||
default_test_content_path = params.default_test_text_path
|
||||
text_test_content = gr.Text(label=i18n("请输入待推理文本路径"), value=default_test_content_path)
|
||||
gr.Markdown(value=i18n("2.3：启动推理服务，如果还没启动的话"))
|
||||
gr.Markdown(value=i18n("2.4：开始批量推理，这个过程比较耗时，可以去干点别的"))
|
||||
with gr.Row():
|
||||
button_model_inference = gr.Button(i18n("开启批量推理"), variant="primary")
|
||||
text_model_inference_info = gr.Text(label=i18n("批量推理结果"), value="", interactive=False)
|
||||
with gr.Tab(label=i18n("第三步：进行参考音频效果校验与筛选"), open=False):
|
||||
gr.Markdown(value=i18n("3.1：启动asr，获取推理音频文本"))
|
||||
default_asr_audio_dir = common.check_path_existence_and_return(
|
||||
os.path.join(default_base_dir, params.inference_audio_dir, params.inference_audio_text_aggregation_dir))
|
||||
text_asr_audio_dir = gr.Text(label=i18n("待asr的音频所在目录"), value=default_asr_audio_dir,
|
||||
interactive=True)
|
||||
with gr.Row():
|
||||
dropdown_asr_model = gr.Dropdown(
|
||||
label=i18n("ASR 模型"),
|
||||
choices=[],
|
||||
interactive=True,
|
||||
value="达摩 ASR (中文)"
|
||||
)
|
||||
dropdown_asr_size = gr.Dropdown(
|
||||
label=i18n("ASR 模型尺寸"),
|
||||
choices=["large"],
|
||||
interactive=True,
|
||||
value="large"
|
||||
)
|
||||
dropdown_asr_lang = gr.Dropdown(
|
||||
label=i18n("ASR 语言设置"),
|
||||
choices=["zh"],
|
||||
interactive=True,
|
||||
value="zh"
|
||||
)
|
||||
with gr.Row():
|
||||
button_asr = gr.Button(i18n("启动asr"), variant="primary")
|
||||
text_asr_info = gr.Text(label=i18n("asr结果"), value="", interactive=False)
|
||||
gr.Markdown(value=i18n("3.2：启动文本相似度分析"))
|
||||
default_text_similarity_analysis_path = common.check_path_existence_and_return(
|
||||
os.path.join(default_base_dir, params.asr_filename + '.list'))
|
||||
text_text_similarity_analysis_path = gr.Text(label=i18n("待分析的文件路径"),
|
||||
value=default_text_similarity_analysis_path, interactive=True)
|
||||
button_asr.click(asr, [text_work_space_dir, text_role, text_asr_audio_dir, dropdown_asr_model,
|
||||
dropdown_asr_size, dropdown_asr_lang],
|
||||
[text_asr_info, text_text_similarity_analysis_path])
|
||||
with gr.Row():
|
||||
button_text_similarity_analysis = gr.Button(i18n("启动文本相似度分析"), variant="primary")
|
||||
text_text_similarity_analysis_info = gr.Text(label=i18n("文本相似度分析结果"), value="",
|
||||
interactive=False)
|
||||
button_text_similarity_analysis.click(text_similarity_analysis, [text_work_space_dir, text_role,
|
||||
text_text_similarity_analysis_path],
|
||||
[text_text_similarity_analysis_info])
|
||||
gr.Markdown(value=i18n("3.3：根据相似度分析结果，重点检查最后几条是否存在复读等问题"))
|
||||
gr.Markdown(value=i18n("3.4：对结果按音频相似度排序，筛选低音质音频"))
|
||||
with gr.Row():
|
||||
text_base_audio_path = gr.Text(label=i18n("请输入基准音频"), value="")
|
||||
text_compare_audio_dir = gr.Text(label=i18n("请输入待比较的音频文件目录"), value="")
|
||||
with gr.Row():
|
||||
button_similarity_audio_output = gr.Button(i18n("输出相似度-参考音频到临时目录"), variant="primary")
|
||||
text_similarity_audio_output_info = gr.Text(label=i18n("输出结果"), value="", interactive=False)
|
||||
button_similarity_audio_output.click(similarity_audio_output,
|
||||
[text_work_space_dir, text_role, text_base_audio_path,
|
||||
text_compare_audio_dir], [text_similarity_audio_output_info])
|
||||
with gr.Row():
|
||||
default_sync_ref_audio_dir = common.check_path_existence_and_return(
|
||||
os.path.join(default_base_dir, params.reference_audio_dir))
|
||||
text_sync_ref_audio_dir = gr.Text(label=i18n("参考音频路径"), value=default_sync_ref_audio_dir,
|
||||
interactive=True)
|
||||
default_sync_inference_audio_dir = common.check_path_existence_and_return(
|
||||
os.path.join(default_base_dir, params.inference_audio_dir))
|
||||
text_sync_inference_audio_dir = gr.Text(label=i18n("被同步的推理音频路径"),
|
||||
value=default_sync_inference_audio_dir, interactive=True)
|
||||
with gr.Row():
|
||||
button_sync_ref_audio = gr.Button(i18n("将参考音频的删除情况，同步到推理音频目录"), variant="primary")
|
||||
text_sync_ref_info = gr.Text(label=i18n("同步结果"), value="", interactive=False)
|
||||
button_sync_ref_audio.click(sync_ref_audio, [text_work_space_dir, text_role, text_sync_ref_audio_dir,
|
||||
text_sync_inference_audio_dir], [text_sync_ref_info])
|
||||
with gr.Tab("第四步：生成参考音频配置文本", open=False):
|
||||
gr.Markdown(value=i18n("4.1：编辑模板"))
|
||||
default_template_path = params.default_template_path
|
||||
default_template_content = common.read_file(default_template_path)
|
||||
text_template_path = gr.Text(label=i18n("模板文件路径"), value=default_template_path, interactive=True)
|
||||
text_template = gr.Text(label=i18n("模板内容"), value=default_template_content, lines=10)
|
||||
gr.Markdown(value=i18n("4.2：生成配置"))
|
||||
default_sync_ref_audio_dir2 = common.check_path_existence_and_return(
|
||||
os.path.join(default_base_dir, params.reference_audio_dir))
|
||||
text_sync_ref_audio_dir2 = gr.Text(label=i18n("参考音频路径"), value=default_sync_ref_audio_dir2,
|
||||
interactive=True)
|
||||
with gr.Row():
|
||||
button_create_config = gr.Button(i18n("生成配置"), variant="primary")
|
||||
text_create_config_info = gr.Text(label=i18n("生成结果"), value="", interactive=False)
|
||||
button_create_config.click(create_config,
|
||||
[text_work_space_dir, text_role, text_template, text_sync_ref_audio_dir2],
|
||||
[text_create_config_info])
|
||||
button_sample.click(sample, [text_work_space_dir, text_role, text_sample_dir, text_base_voice_path,
|
||||
text_subsection_num, text_sample_num, checkbox_similarity_output],
|
||||
[text_sample_info, text_model_inference_voice_dir, text_sync_ref_audio_dir,
|
||||
text_sync_ref_audio_dir2])
|
||||
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])
|
||||
|
||||
app.launch(
|
||||
server_port=9423,
|
||||
quiet=True,
|
||||
)
|
||||
app.launch(
|
||||
server_port=9423,
|
||||
quiet=True,
|
||||
)
|
||||
|
||||
@ -1,12 +1,20 @@
|
||||
import time
|
||||
import os
|
||||
import requests
|
||||
import itertools
|
||||
import multiprocessing
|
||||
from multiprocessing import Pool
|
||||
from concurrent.futures import ProcessPoolExecutor
|
||||
import numpy as np
|
||||
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
|
||||
from Ref_Audio_Selector.config_param.log_config import logger
|
||||
from Ref_Audio_Selector.config_param.log_config import logger, p_logger
|
||||
|
||||
|
||||
# 假设手动指定端口范围为9400-9500
|
||||
available_ports = list(range(9400, 9500))
|
||||
|
||||
class URLComposer:
|
||||
def __init__(self, base_url, emotion_param_name, text_param_name, ref_path_param_name, ref_text_param_name):
|
||||
self.base_url = base_url
|
||||
@ -74,8 +82,24 @@ 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):
|
||||
def generate_audio_files_parallel(url_composer, text_list, emotion_list, output_dir_path, num_processes=None):
|
||||
if num_processes is None:
|
||||
num_processes = multiprocessing.cpu_count()
|
||||
|
||||
num_processes = min(num_processes, len(available_ports)) # 限制进程数不超过可用端口数
|
||||
|
||||
# 将emotion_list均匀分成num_processes个子集
|
||||
emotion_groups = np.array_split(emotion_list, num_processes)
|
||||
|
||||
with ProcessPoolExecutor(max_workers=num_processes) as executor:
|
||||
futures = [executor.submit(generate_audio_files_for_emotion_group, url_composer, text_list, group, output_dir_path)
|
||||
for group in emotion_groups]
|
||||
for future in futures:
|
||||
future.result() # 等待所有进程完成
|
||||
|
||||
|
||||
def generate_audio_files_for_emotion_group(url_composer, text_list, emotion_list, output_dir_path):
|
||||
start_time = time.perf_counter() # 使用 perf_counter 获取高精度计时起点
|
||||
# Ensure the output directory exists
|
||||
output_dir = os.path.abspath(output_dir_path)
|
||||
os.makedirs(output_dir, exist_ok=True)
|
||||
@ -108,7 +132,7 @@ def generate_audio_files(url_composer, text_list, emotion_list, output_dir_path)
|
||||
# 检查是否已经存在对应的音频文件，如果存在则跳过
|
||||
if os.path.exists(text_subdir_text_file_path) and os.path.exists(emotion_subdir_emotion_file_path):
|
||||
has_generated_count += 1
|
||||
logger.info(f"进度: {has_generated_count}/{all_count}")
|
||||
logger.info(f"进程ID: {os.getpid()}, 进度: {has_generated_count}/{all_count}")
|
||||
continue
|
||||
|
||||
if url_composer.is_emotion():
|
||||
@ -126,6 +150,11 @@ def generate_audio_files(url_composer, text_list, emotion_list, output_dir_path)
|
||||
|
||||
has_generated_count += 1
|
||||
logger.info(f"进度: {has_generated_count}/{all_count}")
|
||||
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"进程ID: {os.getpid()}, generate_audio_files_for_emotion_group 执行耗时: {elapsed_time:.6f} 秒"
|
||||
p_logger.info(log_message)
|
||||
|
||||
|
||||
def inference_audio_from_api(url):
|
||||
|
||||
Loading…
x
Reference in New Issue
Block a user