Closed wikeeyang closed 1 year ago
以上补充信息,供您参考,请帮忙分析一下,谢谢!
已经通过分别在两台机器上部署ChatGLM-6B、PaddleSpeech,并改用API调用ChatGLM-6B问答的方式解决。
可能是显存不够,我用的是3060 12G的显卡,从chatglm的回复加生成语音的时间首次大概要30多秒以上,chatglm和paddlespeech都挺吃显存的。
这是来自QQ邮箱的假期自动回复邮件。 您好,我最近正在休假中,无法亲自回复您的邮件。我将在假期结束后,尽快给您回复。
环境: Windows 11 x64 NVIDIA P3200 6GB显卡,驱动程序最新版。 32GB 内存 Python 3.9.6 CUDA 11.7/11.8都有 cuDNN 8.4.1/8.5.0/8.8.1/8.9都有,通过改path路径切换 ChatGLM-6b-int4 PaddleSpeech 没有使用anconda,自己通过bat脚本,切换Python环境,机器上有3.9/3.10/3.11多个环境,切换都正常,不会干扰。 之前已经调好对话模型各种环境,试过ChatGLM-6B及int4、闻达RWKV、Replitcode、GPT4ALL等,在多个Python环境中也都没问题,PaddlePaddle2.4.2,在CUDA11.7、cuDNN8.4.1官方要求环境下,手工安装,PaddleSpeech也是手工安装测试,完全正常,根据版主Readme对照,环境基本具备。 依据版主提示进行文字转语音命令行测试,虽有警告信息,但结果正常。如下:(根据警告信息,cuDNN环境升为8.5)。
D:\AITest\PaddleSpeech\ChatGLM6B-Speech>paddlespeech tts --input "今天天气不好,我们暂时就先不约了,等什么时候有时间了,我给你打电话" --output opoint.wav C:\Python\Python39\lib\site-packages\pkg_resources__init.py:121: DeprecationWarning: pkg_resources is deprecated as an API warnings.warn("pkg_resources is deprecated as an API", DeprecationWarning) C:\Python\Python39\lib\site-packages\pkg_resources__init.py:2870: DeprecationWarning: Deprecated call to
pkg_resources.declare_namespace('google')
. Implementing implicit namespace packages (as specified in PEP 420) is preferred topkg_resources.declare_namespace
. See https://setuptools.pypa.io/en/latest/references/keywords.html#keyword-namespace-packages declare_namespace(pkg) C:\Python\Python39\lib\site-packages\pkg_resources__init__.py:2870: DeprecationWarning: Deprecated call topkg_resources.declare_namespace('mpl_toolkits')
. Implementing implicit namespace packages (as specified in PEP 420) is preferred topkg_resources.declare_namespace
. See https://setuptools.pypa.io/en/latest/references/keywords.html#keyword-namespace-packages declare_namespace(pkg) C:\Python\Python39\lib\site-packages\pkg_resources\init__.py:2870: DeprecationWarning: Deprecated call topkg_resources.declare_namespace('google')
. Implementing implicit namespace packages (as specified in PEP 420) is preferred topkg_resources.declare_namespace
. See https://setuptools.pypa.io/en/latest/references/keywords.html#keyword-namespace-packages declare_namespace(pkg) C:\Python\Python39\lib\site-packages\librosa\core\constantq.py:1059: DeprecationWarning:np.complex
is a deprecated alias for the builtincomplex
. To silence this warning, usecomplex
by itself. Doing this will not modify any behavior and is safe. If you specifically wanted the numpy scalar type, usenp.complex128
here. Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations dtype=np.complex, C:\Python\Python39\lib\site-packages_distutils_hack\init__.py:33: UserWarning: Setuptools is replacing distutils. warnings.warn("Setuptools is replacing distutils.") W0509 07:36:24.462970 14632 gpu_resources.cc:61] Please NOTE: device: 0, GPU Compute Capability: 6.1, Driver API Version: 12.0, Runtime API Version: 11.7 W0509 07:36:24.494622 14632 gpu_resources.cc:91] device: 0, cuDNN Version: 8.5. D:\AITest\PaddleSpeech\ChatGLM6B-Speech\opoint.wav针对版主脚本,仅修改ChatGLM-6B-int4模型路径,我将所有模型文件,分类放在硬盘一个目录下。以及Gradio调用的IP地址和端口。 tokenizer = AutoTokenizer.from_pretrained("D:\AITest\Model\chatglm-6b-int4", trust_remote_code=True, revision="") model = AutoModel.from_pretrained("D:\AITest\Model\chatglm-6b-int4", trust_remote_code=True, revision="").half().cuda()
gradio_app.launch(share=False, inbrowser=True, server_name="127.0.0.1", server_port=8888)
其它都未改。
执行及报错信息如下:
W0509 07:36:24.494622 14632 gpu_resources.cc:91] device: 0, cuDNN Version: 8.5. D:\AITest\PaddleSpeech\ChatGLM6B-Speech\opoint.wav D:\AITest\PaddleSpeech\ChatGLM6B-Speech>python mywebui.py C:\Python\Python39\lib\site-packages\librosa\core\constantq.py:1059: DeprecationWarning:
np.complex
is a deprecated alias for the builtincomplex
. To silence this warning, usecomplex
by itself. Doing this will not modify any behavior and is safe. If you specifically wanted the numpy scalar type, usenp.complex128
here. Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations dtype=np.complex, C:\Python\Python39\lib\site-packages_distutils_hack__init__.py:33: UserWarning: Setuptools is replacing distutils. warnings.warn("Setuptools is replacing distutils.") No compiled kernel found. Compiling kernels : C:\Users\Administrator.cache\huggingface\modules\transformers_modules\chatglm-6b-int4\quantization_kernels_parallel.c Compiling gcc -O3 -fPIC -pthread -fopenmp -std=c99 C:\Users\Administrator.cache\huggingface\modules\transformers_modules\chatglm-6b-int4\quantization_kernels_parallel.c -shared -o C:\Users\Administrator.cache\huggingface\modules\transformers_modules\chatglm-6b-int4\quantization_kernels_parallel.so Load kernel : C:\Users\Administrator.cache\huggingface\modules\transformers_modules\chatglm-6b-int4\quantization_kernels_parallel.so Setting CPU quantization kernel threads to 6 Using quantization cache Applying quantization to glm layers Running on local URL: http://127.0.0.1:8888To create a public link, set
share=True
inlaunch()
. [2023-05-09 08:20:53,875] [ WARNING] logging.py:295 - The dtype of attention mask (torch.int64) is not bool [2023-05-09 08:21:17,522] [ INFO] - Already cached C:\Users\Administrator.paddlenlp\models\bert-base-chinese\bert-base-chinese-vocab.txt [2023-05-09 08:21:17,539] [ INFO] - tokenizer config file saved in C:\Users\Administrator.paddlenlp\models\bert-base-chinese\tokenizer_config.json [2023-05-09 08:21:17,539] [ INFO] - Special tokens file saved in C:\Users\Administrator.paddlenlp\models\bert-base-chinese\special_tokens_map.json W0509 08:21:20.251111 16368 gpu_resources.cc:61] Please NOTE: device: 0, GPU Compute Capability: 6.1, Driver API Version: 12.0, Runtime API Version: 11.7 W0509 08:21:20.261677 16368 gpu_resources.cc:91] device: 0, cuDNN Version: 8.5.D:\AITest\PaddleSpeech\ChatGLM6B-Speech>
屏幕截图: 运行中: 报错后:
烦请版主帮我分析分析可能出错的原因,谢谢!