‘status’ was not declared in this scope; did you mean ‘std::filesystem::status’?

[yourengod]

您好.

/MeloTTS.cpp/src/utils.cpp: In function ‘void SetOneDNN_CPU_MAX_ISA()’:
/MeloTTS.cpp/src/utils.cpp:40:9: error: ‘status’ was not declared in this scope; did you mean ‘std::filesystem::status’?
   40 |     if (status == 0) {
      |         ^~~~~~
      |         std::filesystem::status
In file included from /usr/include/c++/13/filesystem:51,
                 from /home/llm/workspace2/tts/mymelottscpp/MeloTTS.cpp/src/utils.h:11,
                 from /home/llm/workspace2/tts/mymelottscpp/MeloTTS.cpp/src/utils.cpp:9:
/usr/include/c++/13/bits/fs_ops.h:360:15: note: ‘std::filesystem::status’ declared here
  360 |   file_status status(const path& __p, error_code& __ec) noexcept;
      |               ^~~~~~
gmake[2]: *** [CMakeFiles/meloTTS_ov.dir/build.make:118：CMakeFiles/meloTTS_ov.dir/src/utils.cpp.o] 错误 1
gmake[1]: *** [CMakeFiles/Makefile2:83：CMakeFiles/meloTTS_ov.dir/all] 错误 2
gmake: *** [Makefile:91：all] 错误 2

什么原因？

[apinge]

请问是哪个branch？

[yourengod]

请问是哪个branch？

git lfs install
git clone https://github.com/apinge/MeloTTS.cpp.git
这个。

[yourengod]

MeloTTS.cpp-multilang-develop的报这个问题：
./build/meloTTS_ov --model_dir ov_models --input_file inputs_zh.txt --output_filename audio --language ZH

[INFO] Set ONEDNN_PRIMITIVE_CACHE_CAPACITY: 100
[INFO] Set ONEDNN_MAX_CPU_ISA: AVX2_VNNI
INTEL_OPENVINO_DIR: /mnt/mydisk/workspace2/tts/mymelottscpp/openvino_genai_ubuntu22_2024.6.0.0_x86_64
[INFO] OpenVinoTokenizer Initialized!
Set CPU_RUNTIME_CACHE_CAPACITY 0
compile model ov_models/tts_zn_mix_en.xml on CPU using 303ms.
OpenVINO:OpenVINO Runtime
Version : 2024.6.0
Build : 2024.6.0-17404-4c0f47d2335-releases/2024/6

Model Device info:openvino_intel_cpu_plugin
Version : 2024.6.0
Build : 2024.6.0-17404-4c0f47d2335-releases/2024/6

CMUDict::CMUDict: Construct CMUDict
init cppjieba
load opencpop-strict.txt to pinyin_to_symbol_map, containing 429 keys
cppinyin::PinyinEncoder Constructed!
[INFO] TextNormalizer is constructed!
[INFO] Init Chinese language Module Succeed!
Set CPU_RUNTIME_CACHE_CAPACITY 0
terminate called after throwing an instance of 'ov::Exception'
what(): Exception from src/inference/src/cpp/core.cpp:121:
Exception from src/frontends/ir/src/ir_deserializer.cpp:355:
Incorrect weights in bin file!

[apinge]

请问是哪个branch？

git lfs install git clone https://github.com/apinge/MeloTTS.cpp.git 这个。

这个branch 我先修一下

[yourengod]

请问是哪个branch？

git lfs install git clone https://github.com/apinge/MeloTTS.cpp.git 这个。

这个branch 我先修一下
谢谢，这个问题是怎么一回事啊？
#28 (comment)

[apinge]

MeloTTS.cpp-multilang-develop的报这个问题： ./build/meloTTS_ov --model_dir ov_models --input_file inputs_zh.txt --output_filename audio --language ZH

[INFO] Set ONEDNN_PRIMITIVE_CACHE_CAPACITY: 100 [INFO] Set ONEDNN_MAX_CPU_ISA: AVX2_VNNI INTEL_OPENVINO_DIR: /mnt/mydisk/workspace2/tts/mymelottscpp/openvino_genai_ubuntu22_2024.6.0.0_x86_64 [INFO] OpenVinoTokenizer Initialized! Set CPU_RUNTIME_CACHE_CAPACITY 0 compile model ov_models/tts_zn_mix_en.xml on CPU using 303ms. OpenVINO:OpenVINO Runtime Version : 2024.6.0 Build : 2024.6.0-17404-4c0f47d2335-releases/2024/6

Model Device info:openvino_intel_cpu_plugin Version : 2024.6.0 Build : 2024.6.0-17404-4c0f47d2335-releases/2024/6

CMUDict::CMUDict: Construct CMUDict init cppjieba load opencpop-strict.txt to pinyin_to_symbol_map, containing 429 keys cppinyin::PinyinEncoder Constructed! [INFO] TextNormalizer is constructed! [INFO] Init Chinese language Module Succeed! Set CPU_RUNTIME_CACHE_CAPACITY 0 terminate called after throwing an instance of 'ov::Exception' what(): Exception from src/inference/src/cpp/core.cpp:121: Exception from src/frontends/ir/src/ir_deserializer.cpp:355: Incorrect weights in bin file!

你这个问题看上去像是BERT的IR 可能是/ov_models/bert_ZH_int8.bin 没有正确的load上来
可能git lfs没搞对，模型没下完全。
列一下 我本地的部分IR文件大小，尤其注意*.bin 文件大小，是不是与你下载的差不多

92M bert_EN_int8.bin
576K bert_EN_int8.xml
147M bert_ZH_int8.bin
576K bert_ZH_int8.xml
576K bert_ZH_static_int8.xml
81M tts_en.bin
41M tts_en_int8.bin
4.1M tts_en_int8.xml
3.3M tts_en.xml
81M tts_zn_mix_en.bin
41M tts_zn_mix_en_int8.bin
3.7M tts_zn_mix_en_int8.xml
3.3M tts_zn_mix_en.xml

[apinge]

您好.

/MeloTTS.cpp/src/utils.cpp: In function ‘void SetOneDNN_CPU_MAX_ISA()’:
/MeloTTS.cpp/src/utils.cpp:40:9: error: ‘status’ was not declared in this scope; did you mean ‘std::filesystem::status’?
   40 |     if (status == 0) {
      |         ^~~~~~
      |         std::filesystem::status
In file included from /usr/include/c++/13/filesystem:51,
                 from /home/llm/workspace2/tts/mymelottscpp/MeloTTS.cpp/src/utils.h:11,
                 from /home/llm/workspace2/tts/mymelottscpp/MeloTTS.cpp/src/utils.cpp:9:
/usr/include/c++/13/bits/fs_ops.h:360:15: note: ‘std::filesystem::status’ declared here
  360 |   file_status status(const path& __p, error_code& __ec) noexcept;
      |               ^~~~~~
gmake[2]: *** [CMakeFiles/meloTTS_ov.dir/build.make:118：CMakeFiles/meloTTS_ov.dir/src/utils.cpp.o] 错误 1
gmake[1]: *** [CMakeFiles/Makefile2:83：CMakeFiles/meloTTS_ov.dir/all] 错误 2
gmake: *** [Makefile:91：all] 错误 2

什么原因？
修好了
be7f064

[yourengod]

您好，问一下，1.说话者有多个中英文的吗，可以设定吗？2.音量可以设定吗？3.流式可以处理吗？4.变成websocket这种多进程有没有问题，可以吧？

[apinge]

1. 中英文需要clone https://github.com/apinge/MeloTTS.cpp/tree/multilang-develop 这个branch，ZH_MIX_EN（default branch）只有中文（中文里可以夹杂英文 请看这个例子https://github.com/apinge/MeloTTS.cpp/blob/ZH_MIX_EN/inputs.txt）
2. 模型本身没有设定音量的输入 可以通过 https://github.com/apinge/MeloTTS.cpp/blob/ZH_MIX_EN/src/tts.h#L58 这样的函数设定
3. 流式这个repo里没做，但可以做的
4. 多进程没试过 理论上可以的吧

[yourengod]

1. 中英文需要clone https://github.com/apinge/MeloTTS.cpp/tree/multilang-develop 这个branch，ZH_MIX_EN（default branch）只有中文（中文里可以夹杂英文 请看这个例子https://github.com/apinge/MeloTTS.cpp/blob/ZH_MIX_EN/inputs.txt）
2. 模型本身没有设定音量的输入 可以通过 https://github.com/apinge/MeloTTS.cpp/blob/ZH_MIX_EN/src/tts.h#L58 这样的函数设定
3. 流式这个repo里没做，但可以做的
4. 多进程没试过 理论上可以的吧

流式怎么做的？

[apinge]

1. 中英文需要clone https://github.com/apinge/MeloTTS.cpp/tree/multilang-develop 这个branch，ZH_MIX_EN（default branch）只有中文（中文里可以夹杂英文 请看这个例子https://github.com/apinge/MeloTTS.cpp/blob/ZH_MIX_EN/inputs.txt）
2. 模型本身没有设定音量的输入 可以通过 https://github.com/apinge/MeloTTS.cpp/blob/ZH_MIX_EN/src/tts.h#L58 这样的函数设定
3. 流式这个repo里没做，但可以做的
4. 多进程没试过 理论上可以的吧

流式怎么做的？

具体细节我不是很清楚
本身一大段文字进来，这个pipeline也会分句推理
大约是搞个 queue的结构 First In First Out 往里输入string