Created by: fengyunshuo

第一步：使用docker安装paddle docker pull hub.baidubce.com/paddlepaddle/paddle:1.6.0-gpu-cuda10.0-cudnn7

docker run --runtime nvidia --name paddle -p 8129:8888 -it -v $PWD:/paddle hub.baidubce.com/paddlepaddle/paddle:1.6.0-gpu-cuda10.0-cudnn7 /bin/bash

在容器内使用python解释器检查显示Your Paddle Fluid is installed succesfully!

第二步：设置变量以及安装所需包 export CUDA_VISIBLE_DEVICES=0

pip3 install zmq

export PYTHONPATH=./:$PYTHONPATH

解压中文与训练模型

等等等

第三步：开启向量服务器

python3 ernie/service/encoder_server.py -m ./inf_model/ -p 8888 -v --encode_layer pooler

显示如下： [INFO] 2019-11-06 03:03:51,781 [encoder_server.py: 65]: propeller server listent on port 8888 [INFO] 2019-11-06 03:03:51,783 [ server.py: 131]: InferenceProxy starting... [DEBUG] 2019-11-06 03:03:51,784 [ server.py: 65]: run_worker 0 [DEBUG] 2019-11-06 03:03:51,785 [ server.py: 68]: cuda_env 0 [INFO] 2019-11-06 03:03:51,787 [ server.py: 140]: Queue init done [DEBUG] 2019-11-06 03:03:53,028 [ server.py: 76]: Predictor building 0 [DEBUG] 2019-11-06 03:03:53,028 [ server.py: 50]: create predictor on card 0 I1106 03:03:53.876863 559 analysis_predictor.cc:88] Profiler is deactivated, and no profiling report will be generated. I1106 03:03:53.893898 559 op_compatible_info.cc:201] The default operator required version is missing. Please update the model version. I1106 03:03:53.893931 559 analysis_predictor.cc:847] MODEL VERSION: 0.0.0 I1106 03:03:53.893942 559 analysis_predictor.cc:849] PREDICTOR VERSION: 1.6.0 --- Running analysis [ir_graph_build_pass] --- Running analysis [ir_graph_clean_pass] --- Running analysis [ir_analysis_pass] --- Running IR pass [is_test_pass] --- Running IR pass [simplify_with_basic_ops_pass] --- Running IR pass [conv_affine_channel_fuse_pass] --- Running IR pass [conv_eltwiseadd_affine_channel_fuse_pass] --- Running IR pass [conv_bn_fuse_pass] --- Running IR pass [conv_eltwiseadd_bn_fuse_pass] --- Running IR pass [multihead_matmul_fuse_pass] I1106 03:03:55.595218 559 graph_pattern_detector.cc:96] --- detected 12 subgraphs --- Running IR pass [fc_fuse_pass] I1106 03:03:55.610505 559 graph_pattern_detector.cc:96] --- detected 12 subgraphs I1106 03:03:55.618679 559 graph_pattern_detector.cc:96] --- detected 25 subgraphs --- Running IR pass [fc_elementwise_layernorm_fuse_pass] I1106 03:03:55.633041 559 graph_pattern_detector.cc:96] --- detected 24 subgraphs --- Running IR pass [conv_elementwise_add_act_fuse_pass] --- Running IR pass [conv_elementwise_add2_act_fuse_pass] --- Running IR pass [conv_elementwise_add_fuse_pass] --- Running IR pass [transpose_flatten_concat_fuse_pass] --- Running IR pass [runtime_context_cache_pass] --- Running analysis [ir_params_sync_among_devices_pass] I1106 03:03:55.645728 559 ir_params_sync_among_devices_pass.cc:41] Sync params from CPU to GPU --- Running analysis [adjust_cudnn_workspace_size_pass] --- Running analysis [inference_op_replace_pass] --- Running analysis [ir_graph_to_program_pass] I1106 03:03:55.840788 559 analysis_predictor.cc:473] ======= optimize end ======= [DEBUG] 2019-11-06 03:03:55,841 [ server.py: 78]: Predictor 0

应该是正确启动了

第四步：客户端请求

python代码如下： from ernie.service.client import ErnieClient client = ErnieClient('./config/vocab.txt', host='localhost', port=8888) ret = client(['谁有狂三这张高清的', '英雄联盟什么英雄最好'])

客户端报错： [INFO] 2019-11-06 03:10:14,655 [ client.py: 60]: Connecting to server... tcp://localhost:8888 [ERROR] 2019-11-06 03:10:24,685 [ client.py: 91]: Resource temporarily unavailable Traceback (most recent call last): File "/home/ERNIE/propeller/service/client.py", line 88, in get reply = await socket.recv(zmq.NOBLOCK) File "/usr/lib/python3.5/asyncio/futures.py", line 363, in iter return self.result() # May raise too. File "/usr/lib/python3.5/asyncio/futures.py", line 274, in result raise self._exception File "/usr/local/lib/python3.5/dist-packages/zmq/_future.py", line 315, in _add_recv_event r = recv(**kwargs) File "zmq/backend/cython/socket.pyx", line 796, in zmq.backend.cython.socket.Socket.recv File "zmq/backend/cython/socket.pyx", line 832, in zmq.backend.cython.socket.Socket.recv File "zmq/backend/cython/socket.pyx", line 191, in zmq.backend.cython.socket._recv_copy File "zmq/backend/cython/socket.pyx", line 186, in zmq.backend.cython.socket._recv_copy File "zmq/backend/cython/checkrc.pxd", line 19, in zmq.backend.cython.checkrc._check_rc zmq.error.Again: Resource temporarily unavailable Traceback (most recent call last): File "test_client.py", line 3, in ret = client(['\u8c01\u6709\u72c2\u4e09\u8fd9\u5f20\u9ad8\u6e05\u7684', '\u82f1\u96c4\u8054\u76df\u4ec0\u4e48\u82f1\u96c4\u6700\u597d']) File "/home/ERNIE/ernie/service/client.py", line 73, in call ret, = super(ErnieClient, self).call(sen_ids, token_type_ids) File "/home/ERNIE/propeller/service/client.py", line 102, in call raise RuntimeError('Client call failed') RuntimeError: Client call failed

服务端报错： W1106 03:21:16.098963 1561 device_context.cc:235] Please NOTE: device: 0, CUDA Capability: 70, Driver API Version: 10.1, Runtime API Version: 10.0 W1106 03:21:16.105224 1561 device_context.cc:243] device: 0, cuDNN Version: 7.5. W1106 03:21:19.452029 1561 naive_executor.cc:43] The NaiveExecutor can not work properly if the cmake flag ON_INFER is not set. W1106 03:21:19.452064 1561 naive_executor.cc:45] Unlike the training phase, all the scopes and variables will be reused to save the allocation overhead. W1106 03:21:19.452075 1561 naive_executor.cc:48] Please re-compile the inference library by setting the cmake flag ON_INFER=ON if you are running Paddle Inference F1106 03:21:20.245061 1561 device_context.cc:318] cudaStreamSynchronize unspecified launch failure errno: 4

一堆以下内容： Exception: /paddle/paddle/fluid/operators/lookup_table_op.cu:43 Assertion id < N failed. Variable value (input) of OP(fluid.layers.embedding) expected >= 0 and < 18000, but got 4656722015836110848. Please check input value. Exception: /paddle/paddle/fluid/operators/lookup_table_op.cu:43 Assertion id < N failed. Variable value (input) of OP(fluid.layers.embedding) expected >= 0 and < 18000, but got 4656722015836110848. Please check input value.

*** Check failure stack trace: *** @ 0x7fa91bd2811d google::LogMessage::Fail() @ 0x7fa91bd2bbcc google::LogMessage::SendToLog() @ 0x7fa91bd27c43 google::LogMessage::Flush() @ 0x7fa91bd2d0de google::LogMessageFatal::~LogMessageFatal() @ 0x7fa91e50b587 paddle::platform::CUDADeviceContext::Wait() @ 0x7fa91e4a0163 paddle::framework::TransDataDevice() @ 0x7fa91e49f11e paddle::framework::TransformData() @ 0x7fa91e48212b paddle::framework::OperatorWithKernel::PrepareData() @ 0x7fa91e483508 paddle::framework::OperatorWithKernel::RunImpl() @ 0x7fa91e483af3 paddle::framework::OperatorWithKernel::RunImpl() @ 0x7fa91e47e2b3 paddle::framework::OperatorBase::Run() @ 0x7fa91e438ef0 paddle::framework::NaiveExecutor::Run() @ 0x7fa91be04cac paddle::AnalysisPredictor::Run() @ 0x7fa91bcf7cd6 ZZN8pybind1112cpp_function10initializeIZN6paddle6pybind12_GLOBAL__N_121BindAnalysisPredictorEPNS_6moduleEEUlRNS2_17AnalysisPredictorERKSt6vectorINS2_12PaddleTensorESaISA_EEE_SC_IS8_SE_EINS_4nameENS_9is_methodENS_7siblingEEEEvOT_PFT0_DpT1_EDpRKT2_ENUlRNS_6detail13function_callEE1_4_FUNESW @ 0x7fa91bbfc176 pybind11::cpp_function::dispatcher() @ 0x4e1307 PyCFunction_Call @ 0x535fcb PyEval_EvalFrameEx @ 0x53a81b PyEval_EvalCodeEx @ 0x4e3537 (unknown) @ 0x5c3bd7 PyObject_Call @ 0x532a22 PyEval_EvalFrameEx @ 0x53af6a PyEval_EvalCodeEx @ 0x4e3423 (unknown) @ 0x5c3bd7 PyObject_Call @ 0x4f08be (unknown) @ 0x5c3bd7 PyObject_Call @ 0x55fbf6 (unknown) @ 0x5c3bd7 PyObject_Call @ 0x5354af PyEval_EvalFrameEx @ 0x53af6a PyEval_EvalCodeEx @ 0x4e3537 (unknown) @ 0x5c3bd7 PyObject_Call

请问是我的环境问题吗还是什么问题呢？谢谢