在nvidia Jetson TX2部署, 编译paddleocr c++ 预测demo --use_tensorrt=true的问题

[WYQ-Github]

完整报错如下
./build/ppocr --limit_side_len=960 --visualize=true --precision=fp32 --gpu_mem=400 --use_tensorrt=true --visualize=true
In PP-OCRv3, rec_image_shape parameter defaults to '3, 48, 320',if you are using recognition model with PP-OCRv2 or an older version, please set --rec_image_shape='3,32,320
total images num: 50
WARNING: Logging before InitGoogleLogging() is written to STDERR
I0601 16:09:40.643213 13695 analysis_predictor.cc:881] TensorRT subgraph engine is enabled
--- Running analysis [ir_graph_build_pass]
--- Running analysis [ir_graph_clean_pass]
--- Running analysis [ir_analysis_pass]
--- Running IR pass [adaptive_pool2d_convert_global_pass]
I0601 16:09:40.696204 13695 fuse_pass_base.cc:57] --- detected 8 subgraphs
--- Running IR pass [shuffle_channel_detect_pass]
--- Running IR pass [quant_conv2d_dequant_fuse_pass]
--- Running IR pass [delete_quant_dequant_op_pass]
--- Running IR pass [delete_quant_dequant_filter_op_pass]
--- Running IR pass [delete_weight_dequant_linear_op_pass]
--- Running IR pass [delete_quant_dequant_linear_op_pass]
--- Running IR pass [add_support_int8_pass]
I0601 16:09:41.050448 13695 fuse_pass_base.cc:57] --- detected 237 subgraphs
--- Running IR pass [simplify_with_basic_ops_pass]
--- Running IR pass [embedding_eltwise_layernorm_fuse_pass]
--- Running IR pass [preln_embedding_eltwise_layernorm_fuse_pass]
--- Running IR pass [multihead_matmul_fuse_pass_v2]
--- Running IR pass [multihead_matmul_fuse_pass_v3]
--- Running IR pass [skip_layernorm_fuse_pass]
--- Running IR pass [preln_skip_layernorm_fuse_pass]
--- Running IR pass [conv_bn_fuse_pass]
I0601 16:09:41.226161 13695 fuse_pass_base.cc:57] --- detected 33 subgraphs
--- Running IR pass [unsqueeze2_eltwise_fuse_pass]
--- Running IR pass [trt_squeeze2_matmul_fuse_pass]
--- Running IR pass [trt_reshape2_matmul_fuse_pass]
--- Running IR pass [trt_flatten2_matmul_fuse_pass]
--- Running IR pass [trt_map_matmul_v2_to_mul_pass]
--- Running IR pass [trt_map_matmul_v2_to_matmul_pass]
--- Running IR pass [trt_map_matmul_to_mul_pass]
--- Running IR pass [fc_fuse_pass]
--- Running IR pass [conv_elementwise_add_fuse_pass]
I0601 16:09:41.302793 13695 fuse_pass_base.cc:57] --- detected 49 subgraphs
--- Running IR pass [tensorrt_subgraph_pass]
I0601 16:09:41.380200 13695 tensorrt_subgraph_pass.cc:145] --- detect a sub-graph with 187 nodes
I0601 16:09:41.415360 13695 tensorrt_subgraph_pass.cc:433] Prepare TRT engine (Optimize model structure, Select OP kernel etc). This process may cost a lot of time.
I0601 16:09:43.589931 13695 engine.cc:222] Run Paddle-TRT Dynamic Shape mode.
W0601 16:09:43.592581 13695 helper.h:107] DLA requests all profiles have same min, max, and opt value. All dla layers are falling back to GPU
I0601 16:12:27.171494 13695 engine.cc:462] ====== engine info ======
I0601 16:12:27.190670 13695 engine.cc:467] Layers:
conv2d (Output: batch_norm_0.tmp_315)
PWN(PWN((Unnamed Layer* 1) [Activation]), hard_swish (Output: hardswish_0.tmp_017))
conv2d (Output: batch_norm_1.tmp_330) + relu (Output: relu_0.tmp_032)
conv2d (Output: depthwise_conv2d_0.tmp_035) + batchnorm_add_scale (Output: batch_norm_2.tmp_345) + relu (Output: relu_1.tmp_047)
conv2d (Output: batch_norm_3.tmp_360) + elementwise (Output: elementwise_add_062)
conv2d (Output: batch_norm_4.tmp_375) + relu (Output: relu_2.tmp_077)
conv2d (Output: depthwise_conv2d_1.tmp_080) + batchnorm_add_scale (Output: batch_norm_5.tmp_390) + relu (Output: relu_3.tmp_092)
conv2d (Output: batch_norm_6.tmp_3105)
conv2d (Output: batch_norm_7.tmp_3118) + relu (Output: relu_4.tmp_0120)
conv2d (Output: depthwise_conv2d_2.tmp_0123) + batchnorm_add_scale (Output: batch_norm_8.tmp_3133) + relu (Output: relu_5.tmp_0135)
conv2d (Output: batch_norm_9.tmp_3148) + elementwise (Output: elementwise_add_1150)
conv2d (Output: batch_norm_10.tmp_3163) + relu (Output: relu_6.tmp_0165)
conv2d (Output: conv2d_114.tmp_0775)
pool2d (Output: pool2d_3.tmp_0777)
conv2d (Output: depthwise_conv2d_3.tmp_0168) + batchnorm_add_scale (Output: batch_norm_11.tmp_3178) + relu (Output: relu_7.tmp_0180)
conv2d (Output: conv2d_115.tmp_1783) + relu (Output: relu_15.tmp_0785)
conv2d (Output: conv2d_116.tmp_1791)
conv2d (Output: batch_norm_12.tmp_3193)
conv2d (Output: batch_norm_13.tmp_3206) + relu (Output: relu_8.tmp_0208)
conv2d (Output: depthwise_conv2d_4.tmp_0211) + batchnorm_add_scale (Output: batch_norm_14.tmp_3221) + relu (Output: relu_9.tmp_0223)
conv2d (Output: batch_norm_15.tmp_3236) + elementwise (Output: elementwise_add_2238)
conv2d (Output: batch_norm_16.tmp_3251) + relu (Output: relu_10.tmp_0253)
conv2d (Output: depthwise_conv2d_5.tmp_0256) + batchnorm_add_scale (Output: batch_norm_17.tmp_3266) + relu (Output: relu_11.tmp_0268)
conv2d (Output: batch_norm_18.tmp_3281) + elementwise (Output: elementwise_add_3283)
conv2d (Output: batch_norm_19.tmp_3296)
conv2d (Output: conv2d_111.tmp_0750)
pool2d (Output: pool2d_2.tmp_0752)
PWN(PWN((Unnamed Layer* 53) [Activation]), hard_swish (Output: hardswish_1.tmp_0298))
conv2d (Output: conv2d_112.tmp_1758) + relu (Output: relu_14.tmp_0760)
conv2d (Output: depthwise_conv2d_6.tmp_0301) + batchnorm_add_scale (Output: batch_norm_20.tmp_3311)
conv2d (Output: conv2d_113.tmp_1766)
PWN(PWN((Unnamed Layer* 60) [Activation]), hard_swish (Output: hardswish_2.tmp_0313))
conv2d (Output: batch_norm_21.tmp_3326)
conv2d (Output: batch_norm_22.tmp_3339)
PWN(PWN((Unnamed Layer* 67) [Activation]), hard_swish (Output: hardswish_3.tmp_0341))
conv2d (Output: depthwise_conv2d_7.tmp_0344) + batchnorm_add_scale (Output: batch_norm_23.tmp_3354)
PWN(PWN((Unnamed Layer* 72) [Activation]), hard_swish (Output: hardswish_4.tmp_0356))
conv2d (Output: batch_norm_24.tmp_3369) + elementwise (Output: elementwise_add_4371)
conv2d (Output: batch_norm_25.tmp_3384)
PWN(PWN((Unnamed Layer* 77) [Activation]), hard_swish (Output: hardswish_5.tmp_0386))
conv2d (Output: depthwise_conv2d_8.tmp_0389) + batchnorm_add_scale (Output: batch_norm_26.tmp_3399)
PWN(PWN((Unnamed Layer* 81) [Activation]), hard_swish (Output: hardswish_6.tmp_0401))
conv2d (Output: batch_norm_27.tmp_3414) + elementwise (Output: elementwise_add_5416)
conv2d (Output: batch_norm_28.tmp_3429)
PWN(PWN((Unnamed Layer* 86) [Activation]), hard_swish (Output: hardswish_7.tmp_0431))
conv2d (Output: depthwise_conv2d_9.tmp_0434) + batchnorm_add_scale (Output: batch_norm_29.tmp_3444)
PWN(PWN((Unnamed Layer* 90) [Activation]), hard_swish (Output: hardswish_8.tmp_0446))
conv2d (Output: batch_norm_30.tmp_3459) + elementwise (Output: elementwise_add_6461)
conv2d (Output: batch_norm_31.tmp_3474)
PWN(PWN((Unnamed Layer* 95) [Activation]), hard_swish (Output: hardswish_9.tmp_0476))
conv2d (Output: depthwise_conv2d_10.tmp_0479) + batchnorm_add_scale (Output: batch_norm_32.tmp_3489)
PWN(PWN((Unnamed Layer* 99) [Activation]), hard_swish (Output: hardswish_10.tmp_0491))
conv2d (Output: batch_norm_33.tmp_3504)
conv2d (Output: batch_norm_34.tmp_3517)
PWN(PWN((Unnamed Layer* 103) [Activation]), hard_swish (Output: hardswish_11.tmp_0519))
conv2d (Output: depthwise_conv2d_11.tmp_0522) + batchnorm_add_scale (Output: batch_norm_35.tmp_3532)
PWN(PWN((Unnamed Layer* 107) [Activation]), hard_swish (Output: hardswish_12.tmp_0534))
conv2d (Output: batch_norm_36.tmp_3547) + elementwise (Output: elementwise_add_7549)
conv2d (Output: batch_norm_37.tmp_3562)
conv2d (Output: conv2d_108.tmp_0725)
pool2d (Output: pool2d_1.tmp_0727)
PWN(PWN((Unnamed Layer* 113) [Activation]), hard_swish (Output: hardswish_13.tmp_0564))
conv2d (Output: conv2d_109.tmp_1733) + relu (Output: relu_13.tmp_0735)
conv2d (Output: depthwise_conv2d_12.tmp_0567) + batchnorm_add_scale (Output: batch_norm_38.tmp_3577)
conv2d (Output: conv2d_110.tmp_1741)
PWN(PWN((Unnamed Layer* 120) [Activation]), hard_swish (Output: hardswish_14.tmp_0579))
conv2d (Output: batch_norm_39.tmp_3592)
conv2d (Output: batch_norm_40.tmp_3605)
PWN(PWN((Unnamed Layer* 127) [Activation]), hard_swish (Output: hardswish_15.tmp_0607))
conv2d (Output: depthwise_conv2d_13.tmp_0610) + batchnorm_add_scale (Output: batch_norm_41.tmp_3620)
PWN(PWN((Unnamed Layer* 132) [Activation]), hard_swish (Output: hardswish_16.tmp_0622))
conv2d (Output: batch_norm_42.tmp_3635) + elementwise (Output: elementwise_add_8637)
conv2d (Output: batch_norm_43.tmp_3650)
PWN(PWN((Unnamed Layer* 137) [Activation]), hard_swish (Output: hardswish_17.tmp_0652))
conv2d (Output: depthwise_conv2d_14.tmp_0655) + batchnorm_add_scale (Output: batch_norm_44.tmp_3665)
PWN(PWN((Unnamed Layer* 141) [Activation]), hard_swish (Output: hardswish_18.tmp_0667))
conv2d (Output: batch_norm_45.tmp_3680) + elementwise (Output: elementwise_add_9682)
conv2d (Output: batch_norm_46.tmp_3695)
PWN(PWN((Unnamed Layer* 146) [Activation]), hard_swish (Output: hardswish_19.tmp_0697))
conv2d (Output: conv2d_105.tmp_0700)
pool2d (Output: pool2d_0.tmp_0702)
conv2d (Output: conv2d_106.tmp_1708) + relu (Output: relu_12.tmp_0710)
conv2d (Output: conv2d_107.tmp_1716)
PWN(PWN(PWN(hard_sigmoid (Output: hardsigmoid_0.tmp_0718)), elementwise (Output: tmp_0720)), elementwise (Output: tmp_1722))
nearest_interp_v2 (Output: nearest_interp_v2_0.tmp_0799)
conv2d (Output: conv2d_117.tmp_0812)
PWN(PWN(PWN(PWN(hard_sigmoid (Output: hardsigmoid_1.tmp_0743)), elementwise (Output: tmp_2745)), elementwise (Output: tmp_3747)), elementwise (Output: tmp_8801))
pool2d (Output: pool2d_4.tmp_0814)
nearest_interp_v2 (Output: nearest_interp_v2_1.tmp_0803)
conv2d (Output: conv2d_120.tmp_0837)
conv2d (Output: conv2d_118.tmp_1820) + relu (Output: relu_16.tmp_0822)
PWN(PWN(PWN(PWN(hard_sigmoid (Output: hardsigmoid_2.tmp_0768)), elementwise (Output: tmp_4770)), elementwise (Output: tmp_5772)), elementwise (Output: tmp_9805))
pool2d (Output: pool2d_5.tmp_0839)
nearest_interp_v2 (Output: nearest_interp_v2_2.tmp_0807)
conv2d (Output: conv2d_123.tmp_0862)
conv2d (Output: conv2d_121.tmp_1845) + relu (Output: relu_17.tmp_0847)
conv2d (Output: conv2d_119.tmp_1828)
PWN(PWN(PWN(PWN(hard_sigmoid (Output: hardsigmoid_3.tmp_0793)), elementwise (Output: tmp_6795)), elementwise (Output: tmp_7797)), elementwise (Output: tmp_10809))
pool2d (Output: pool2d_6.tmp_0864)
conv2d (Output: conv2d_126.tmp_0887)
conv2d (Output: conv2d_124.tmp_1870) + relu (Output: relu_18.tmp_0872)
conv2d (Output: conv2d_122.tmp_1853)
pool2d (Output: pool2d_7.tmp_0889)
PWN(PWN(PWN(hard_sigmoid (Output: hardsigmoid_4.tmp_0830)), elementwise (Output: tmp_11832)), elementwise (Output: tmp_12834))
conv2d (Output: conv2d_127.tmp_1895) + relu (Output: relu_19.tmp_0897)
conv2d (Output: conv2d_125.tmp_1878)
nearest_interp_v2 (Output: nearest_interp_v2_3.tmp_0911)
PWN(PWN(PWN(hard_sigmoid (Output: hardsigmoid_5.tmp_0855)), elementwise (Output: tmp_13857)), elementwise (Output: tmp_14859))
conv2d (Output: conv2d_128.tmp_1903)
nearest_interp_v2 (Output: nearest_interp_v2_4.tmp_0913)
PWN(PWN(PWN(hard_sigmoid (Output: hardsigmoid_6.tmp_0880)), elementwise (Output: tmp_15882)), elementwise (Output: tmp_16884))
nearest_interp_v2 (Output: nearest_interp_v2_5.tmp_0915)
PWN(PWN(PWN(hard_sigmoid (Output: hardsigmoid_7.tmp_0905)), elementwise (Output: tmp_17907)), elementwise (Output: tmp_18909))
nearest_interp_v2_3.tmp_0911 copy
nearest_interp_v2_4.tmp_0913 copy
nearest_interp_v2_5.tmp_0915 copy
conv2d (Output: batch_norm_47.tmp_3930) + relu (Output: batch_norm_47.tmp_4932)
conv2d_transpose (Output: conv2d_transpose_4.tmp_0935) + (Unnamed Layer* 201) [Constant] + (Unnamed Layer* 207) [Shuffle] + elementwise (Output: elementwise_add_10.tmp_0938) + batchnorm_add_scale (Output: batch_norm_48.tmp_3948) + relu (Output: batch_norm_48.tmp_4950)
conv2d_transpose (Output: conv2d_transpose_5.tmp_0953) + (Unnamed Layer* 212) [Constant] + (Unnamed Layer* 218) [Shuffle] + elementwise (Output: elementwise_add_11.tmp_0956)
PWN(sigmoid (Output: sigmoid_0.tmp_0958))

Bindings:
x
sigmoid_0.tmp_0958
I0601 16:12:27.191135 13695 engine.cc:469] ====== engine info end ======
--- Running IR pass [conv_bn_fuse_pass]
--- Running IR pass [conv_elementwise_add_act_fuse_pass]
--- Running IR pass [conv_elementwise_add2_act_fuse_pass]
--- Running IR pass [transpose_flatten_concat_fuse_pass]
--- Running analysis [ir_params_sync_among_devices_pass]
I0601 16:12:27.238207 13695 ir_params_sync_among_devices_pass.cc:100] Sync params from CPU to GPU
terminate called after throwing an instance of 'phi::enforce::EnforceNotMet'
what():

---

C++ Traceback (most recent call last):

0 paddle_infer::CreatePredictor(paddle::AnalysisConfig const&)
1 paddle_infer::Predictor::Predictor(paddle::AnalysisConfig const&)
2 std::unique_ptr<paddle::PaddlePredictor, std::default_deletepaddle::PaddlePredictor > paddle::CreatePaddlePredictor<paddle::AnalysisConfig, (paddle::PaddleEngineKind)2>(paddle::AnalysisConfig const&)
3 paddle::AnalysisPredictor::Init(std::shared_ptrpaddle::framework::Scope const&, std::shared_ptrpaddle::framework::ProgramDesc const&)
4 paddle::AnalysisPredictor::PrepareProgram(std::shared_ptrpaddle::framework::ProgramDesc const&)
5 paddle::AnalysisPredictor::OptimizeInferenceProgram()
6 paddle::inference::analysis::Analyzer::RunAnalysis(paddle::inference::analysis::Argument*)
7 paddle::inference::analysis::IrParamsSyncAmongDevicesPass::RunImpl(paddle::inference::analysis::Argument*)
8 paddle::inference::analysis::IrParamsSyncAmongDevicesPass::CopyParamsToGpu(paddle::inference::analysis::Argument*)
9 paddle::framework::TensorCopySync(phi::DenseTensor const&, phi::Place const&, phi::DenseTensor*)
10 phi::DenseTensor::mutable_data(phi::Place const&, paddle::experimental::DataType, unsigned long)
11 paddle::memory::AllocShared(phi::Place const&, unsigned long)
12 paddle::memory::allocation::AllocatorFacade::AllocShared(phi::Place const&, unsigned long)
13 paddle::memory::allocation::AllocatorFacade::Alloc(phi::Place const&, unsigned long)
14 paddle::memory::allocation::StatAllocator::AllocateImpl(unsigned long)
15 paddle::memory::allocation::Allocator::Allocate(unsigned long)
16 paddle::memory::allocation::RetryAllocator::AllocateImpl(unsigned long)
17 paddle::memory::allocation::NaiveBestFitAllocator::AllocateImpl(unsigned long)
18 paddle::memory::legacy::AllocVisitor::result_type paddle::platform::VisitPlacepaddle::memory::legacy::AllocVisitor(phi::Place const&, paddle::memory::legacy::AllocVisitor const&)
19 void* paddle::memory::legacy::Allocphi::GPUPlace(phi::GPUPlace const&, unsigned long)
20 paddle::memory::legacy::GetGPUBuddyAllocator(int)
21 paddle::memory::legacy::GPUBuddyAllocatorList::Get(int)
22 paddle::memory::legacy::GPUBuddyAllocatorList::Get(int)::{lambda()#1}::operator()() const
23 paddle::platform::GpuMaxChunkSize()
24 paddle::platform::GpuMaxAllocSize()
25 phi::enforce::EnforceNotMet::EnforceNotMet(phi::ErrorSummary const&, char const*, int)
26 phi::enforce::GetCurrentTraceBackStringabi:cxx11

---

Error Message Summary:

ResourceExhaustedError: Not enough available GPU memory.
[Hint: Expected available_to_alloc >= alloc_bytes, but received available_to_alloc:318272921 < alloc_bytes:419430400.] (at /home/paddle/data/xly/workspace/23303/Paddle/paddle/fluid/platform/device/gpu/gpu_info.cc:99)

Aborted (core dumped)

[WYQ-Github]

@an1018 哥哥能帮忙看一下吗

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed in 7 days if no further activity occurs. Thank you for your contributions.