Skip to content

D
DeepSpeech

PaddlePaddle / DeepSpeech
1 年多 前同步成功

通知 207
Star 8425
Fork 1598
D
DeepSpeech
开放中
Opened by saxon_zh@saxon_zhOwner

模型部署后，对长语音是否支持？

Created by: bigcash

您好，感谢开源！ 我部署启动了deploy/demo_server.py，测试大概40分钟的语音，wav文件8000rate的有38M的时候，报内存溢出错误：Out of memory error on GPU 0. 我使用的是2080Ti的单卡进行测试的，显存11019MiB，是不是这种长语音需要对音频进行切割，一般怎样切割可以使最终结果更加准确，是固定时长切割（如按每隔10s切割）还是按录音中没有语音信号时切割的？ 我也使用过百度云中提供的api进行语音转文字，是将40分钟录音文件直接提交的，是不是在服务端先进行了语音分割，然后再对所有结果合并，最后给出结果的？ 多谢解惑！！

server端报错如下（使用BaiduCN1.2k模型）： Received utterance[length=40108716] from 172.16.1.18, saved to demo_cache/20200109031539_172.16.1.18.wav. finish initing model from pretrained params from models/BaiduCN1.2k/ W0109 11:16:28.406390 11887 operator.cc:179] mul raises an exception paddle::memory::allocation::BadAlloc,

C++ Call Stacks (More useful to developers):

0 paddle::memory::detail::GPUAllocator::Alloc(unsigned long*, unsigned long) 1 paddle::memory::detail::BuddyAllocator::RefillPool(unsigned long) 2 paddle::memory::detail::BuddyAllocator::Alloc(unsigned long) 3 void* paddle::memory::legacy::Allocpaddle::platform::CUDAPlace(paddle::platform::CUDAPlace const&, unsigned long) 4 paddle::memory::allocation::NaiveBestFitAllocator::AllocateImpl(unsigned long) 5 paddle::memory::allocation::Allocator::Allocate(unsigned long) 6 paddle::memory::allocation::RetryAllocator::AllocateImpl(unsigned long) 7 paddle::memory::allocation::AllocatorFacade::Alloc(paddle::platform::Place const&, unsigned long) 8 paddle::memory::allocation::AllocatorFacade::AllocShared(paddle::platform::Place const&, unsigned long) 9 paddle::memory::AllocShared(paddle::platform::Place const&, unsigned long) 10 paddle::framework::Tensor::mutable_data(paddle::platform::Place, paddle::framework::proto::VarType_Type, unsigned long) 11 paddle::operators::MulKernel<paddle::platform::CUDADeviceContext, float>::Compute(paddle::framework::ExecutionContext const&) const 12 std::_Function_handler<void (paddle::framework::ExecutionContext const&), paddle::framework::OpKernelRegistrarFunctor<paddle::platform::CUDAPlace, false, 0ul, paddle::operators::MulKernel<paddle::platform::CUDADeviceContext, float>, paddle::operators::MulKernel<paddle::platform::CUDADeviceContext, double>, paddle::operators::MulKernel<paddle::platform::CUDADeviceContext, paddle::platform::float16> >::operator()(char const*, char const*, int) const::{lambda(paddle::framework::ExecutionContext const&)#1 (closed)}>::_M_invoke(std::_Any_data const&, paddle::framework::ExecutionContext const&) 13 paddle::framework::OperatorWithKernel::RunImpl(paddle::framework::Scope const&, paddle::platform::Place const&, paddle::framework::RuntimeContext*) const 14 paddle::framework::OperatorWithKernel::RunImpl(paddle::framework::Scope const&, paddle::platform::Place const&) const 15 paddle::framework::OperatorBase::Run(paddle::framework::Scope const&, paddle::platform::Place const&) 16 paddle::framework::Executor::RunPreparedContext(paddle::framework::ExecutorPrepareContext*, paddle::framework::Scope*, bool, bool, bool) 17 paddle::framework::Executor::Run(paddle::framework::ProgramDesc const&, paddle::framework::Scope*, int, bool, bool, std::vector<std::string, std::allocatorstd::string > const&, bool)

Error Message Summary:

Out of memory error on GPU 0. Cannot allocate 1.912537GB memory on GPU 0, available memory is only 381.125000MB.

Please check whether there is any other process using GPU 0.

  1. If yes, please stop them, or start PaddlePaddle on another GPU.
  2. If no, please try one of the following suggestions:
    1. Decrease the batch size of your model.
    2. FLAGS_fraction_of_gpu_memory_to_use is 0.92 now, please set it to a higher value but less than 1.0. The command is export FLAGS_fraction_of_gpu_memory_to_use=xxx.

at (/paddle/paddle/fluid/memory/detail/system_allocator.cc:151)

Exception happened during processing of request from ('172.16.1.18', 55944) Traceback (most recent call last): File "/home/lingbao/soft/anaconda3/envs/baidu27/lib/python2.7/SocketServer.py", line 293, in _handle_request_noblock self.process_request(request, client_address) File "/home/lingbao/soft/anaconda3/envs/baidu27/lib/python2.7/SocketServer.py", line 321, in process_request self.finish_request(request, client_address) File "/home/lingbao/soft/anaconda3/envs/baidu27/lib/python2.7/SocketServer.py", line 334, in finish_request self.RequestHandlerClass(request, client_address, self) File "/home/lingbao/soft/anaconda3/envs/baidu27/lib/python2.7/SocketServer.py", line 655, in init self.handle() File "deploy/demo_server.py", line 101, in handle transcript = self.server.audio_process_handler(filename) File "deploy/demo_server.py", line 195, in file_to_transcript feeding_dict=data_generator.feeding) File "/home/lingbao/work/text/DeepSpeech/deploy/../model_utils/model.py", line 424, in infer_batch_probs return_numpy=False) File "/home/lingbao/soft/anaconda3/envs/baidu27/lib/python2.7/site-packages/paddle/fluid/executor.py", line 780, in run six.reraise(*sys.exc_info()) File "/home/lingbao/soft/anaconda3/envs/baidu27/lib/python2.7/site-packages/paddle/fluid/executor.py", line 775, in run use_program_cache=use_program_cache) File "/home/lingbao/soft/anaconda3/envs/baidu27/lib/python2.7/site-packages/paddle/fluid/executor.py", line 822, in _run_impl use_program_cache=use_program_cache) File "/home/lingbao/soft/anaconda3/envs/baidu27/lib/python2.7/site-packages/paddle/fluid/executor.py", line 899, in _run_program fetch_var_name) RuntimeError:

C++ Call Stacks (More useful to developers):

0 paddle::memory::detail::GPUAllocator::Alloc(unsigned long*, unsigned long) 1 paddle::memory::detail::BuddyAllocator::RefillPool(unsigned long) 2 paddle::memory::detail::BuddyAllocator::Alloc(unsigned long) 3 void* paddle::memory::legacy::Allocpaddle::platform::CUDAPlace(paddle::platform::CUDAPlace const&, unsigned long) 4 paddle::memory::allocation::NaiveBestFitAllocator::AllocateImpl(unsigned long) 5 paddle::memory::allocation::Allocator::Allocate(unsigned long) 6 paddle::memory::allocation::RetryAllocator::AllocateImpl(unsigned long) 7 paddle::memory::allocation::AllocatorFacade::Alloc(paddle::platform::Place const&, unsigned long) 8 paddle::memory::allocation::AllocatorFacade::AllocShared(paddle::platform::Place const&, unsigned long) 9 paddle::memory::AllocShared(paddle::platform::Place const&, unsigned long) 10 paddle::framework::Tensor::mutable_data(paddle::platform::Place, paddle::framework::proto::VarType_Type, unsigned long) 11 paddle::operators::MulKernel<paddle::platform::CUDADeviceContext, float>::Compute(paddle::framework::ExecutionContext const&) const 12 std::_Function_handler<void (paddle::framework::ExecutionContext const&), paddle::framework::OpKernelRegistrarFunctor<paddle::platform::CUDAPlace, false, 0ul, paddle::operators::MulKernel<paddle::platform::CUDADeviceContext, float>, paddle::operators::MulKernel<paddle::platform::CUDADeviceContext, double>, paddle::operators::MulKernel<paddle::platform::CUDADeviceContext, paddle::platform::float16> >::operator()(char const*, char const*, int) const::{lambda(paddle::framework::ExecutionContext const&)#1 (closed)}>::_M_invoke(std::_Any_data const&, paddle::framework::ExecutionContext const&) 13 paddle::framework::OperatorWithKernel::RunImpl(paddle::framework::Scope const&, paddle::platform::Place const&, paddle::framework::RuntimeContext*) const 14 paddle::framework::OperatorWithKernel::RunImpl(paddle::framework::Scope const&, paddle::platform::Place const&) const 15 paddle::framework::OperatorBase::Run(paddle::framework::Scope const&, paddle::platform::Place const&) 16 paddle::framework::Executor::RunPreparedContext(paddle::framework::ExecutorPrepareContext*, paddle::framework::Scope*, bool, bool, bool) 17 paddle::framework::Executor::Run(paddle::framework::ProgramDesc const&, paddle::framework::Scope*, int, bool, bool, std::vector<std::string, std::allocatorstd::string > const&, bool)

Error Message Summary:

Out of memory error on GPU 0. Cannot allocate 1.912537GB memory on GPU 0, available memory is only 381.125000MB.

Please check whether there is any other process using GPU 0.

  1. If yes, please stop them, or start PaddlePaddle on another GPU.
  2. If no, please try one of the following suggestions:
    1. Decrease the batch size of your model.
    2. FLAGS_fraction_of_gpu_memory_to_use is 0.92 now, please set it to a higher value but less than 1.0. The command is export FLAGS_fraction_of_gpu_memory_to_use=xxx.

at (/paddle/paddle/fluid/memory/detail/system_allocator.cc:151)