search

JDAI-CV / DNNLibrary

Daquexian's NNAPI Library. ONNX + Android NNAPI
Apache License 2.0
346 stars 59 forks source link

onnx2daq fails with yolov3 network #64

Open amelia808 opened 5 years ago

amelia808 commented 5 years ago 
terminate called after throwing an instance of 'std::out_of_range'
  what():  _Map_base::at

this is the error while trying to convert https://github.com/onnx/models/tree/master/vision/object_detection_segmentation/yolov3

amelia808 commented 5 years ago

It actually fails for all onnx object detection models.

ardiya commented 4 years ago

I got similar problem with my own .onnx.

I compiled the code with debug mode and tried to convert ONNX yolov3.onnx and I got this error

#0  __GI_raise (sig=sig@entry=6) at ../sysdeps/unix/sysv/linux/raise.c:50
#1  0x00007ffff7a4c899 in __GI_abort () at abort.c:79
#2  0x00007ffff7e1f5f6 in ?? () from /lib/x86_64-linux-gnu/libstdc++.so.6
#3  0x00007ffff7e2b9ec in ?? () from /lib/x86_64-linux-gnu/libstdc++.so.6
#4  0x00007ffff7e2ba47 in std::terminate() () from /lib/x86_64-linux-gnu/libstdc++.so.6
#5  0x00007ffff7e2bca9 in __cxa_throw () from /lib/x86_64-linux-gnu/libstdc++.so.6
#6  0x00007ffff7e21f04 in std::__throw_out_of_range(char const*) ()
   from /lib/x86_64-linux-gnu/libstdc++.so.6
#7  0x0000555555686023 in std::__detail::_Map_base<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >, std::pair<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const, onnx_daq::Value*>, std::allocator<std::pair<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const, onnx_daq::Value*> >, std::__detail::_Select1st, std::equal_to<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > >, std::hash<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > >, std::__detail::_Mod_range_hashing, std::__detail::_Default_ranged_hash, std::__detail::_Prime_rehash_policy, std::__detail::_Hashtable_traits<true, false, true>, true>::at (this=0x7fffffffc600, __k="W74")
    at /usr/include/c++/9/bits/hashtable_policy.h:750
#8  0x0000555555683999 in std::unordered_map<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >, onnx_daq::Value*, std::hash<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > >, std::equal_to<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > >, std::allocator<std::pair<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const, onnx_daq::Value*> > >::at (this=0x7fffffffc600, __k="W74")
    at /usr/include/c++/9/bits/unordered_map.h:1002
#9  0x000055555567dcaf in onnx_daq::graphProtoToGraph (gp=..., nested=false)
    at /home/ardiya/Workspace/DNNLibrary/third_party/onnx/onnx/common/ir_pb_converter.cc:287
#10 0x000055555567e52d in onnx_daq::ImportModelProto (mp=...)
    at /home/ardiya/Workspace/DNNLibrary/third_party/onnx/onnx/common/ir_pb_converter.cc:331
#11 0x0000555555646fd5 in onnx_daq::optimization::Optimizer::optimize (this=0x7fffffffcb40, mp_in=...)
    at /home/ardiya/Workspace/DNNLibrary/third_party/onnx/onnx/optimizer/optimize.h:26
#12 0x0000555555639d9a in onnx_daq::optimization::OptimizeFixed (mp_in=..., 
    names=std::vector of length 18, capacity 18 = {...})
    at /home/ardiya/Workspace/DNNLibrary/third_party/onnx/onnx/optimizer/optimize.cc:38
#13 0x000055555557711d in dnn::OnnxConverter::Convert (this=0x7fffffffd880, model_proto=..., 
    table_file="") at /home/ardiya/Workspace/DNNLibrary/tools/onnx2daq/OnnxConverter.cpp:646
#14 0x000055555556bfc0 in main (argc=3, argv=0x7fffffffdbc8)
    at /home/ardiya/Workspace/DNNLibrary/tools/onnx2daq/onnx2daq.cpp:34

from frame 9, file ir_pb_converter.cc line 287, which is n->addInput(value_by_name_of.at(input)); here are more info from the gdb

(gdb) print input
$1 = "W74"
(gdb) print value_by_name_of
$2 = std::unordered_map with 478 elements = {["yolonms_layer_1/ExpandDims_1:0"] = 0x55555b61a6e0, 
  ["TFNodes/yolo_evaluation_layer_1/concat_7:0"] = 0x55555b619640, 
  ["TFNodes/yolo_evaluation_layer_1/strided_slice_56:0"] = 0x55555b619140, 
  ["TFNodes/yolo_evaluation_layer_1/add_5:0"] = 0x55555b618880, 
  ["TFNodes/yolo_evaluation_layer_1/strided_slice_54:0"] = 0x55555b618380, 
  ["TFNodes/yolo_evaluation_layer_1/mul_15:0"] = 0x55555b617650, 
  ["TFNodes/yolo_evaluation_layer_1/strided_slice_52:0"] = 0x55555b616d00, 
  ["TFNodes/yolo_evaluation_layer_1/add_4:0"] = 0x55555b616560, 
  ["TFNodes/yolo_evaluation_layer_1/Sigmoid_6:0"] = 0x55555b616310, 
  ["TFNodes/yolo_evaluation_layer_1/strided_slice_46:0"] = 0x55555b615cc0, 
  ["TFNodes/yolo_evaluation_layer_1/truediv_23:0"] = 0x55555b6158a0, 
  ["TFNodes/yolo_evaluation_layer_1/strided_slice_53:0"] = 0x55555b615040, 
  ["TFNodes/yolo_evaluation_layer_1/mul_13:0"] = 0x55555b614a50, 
  ["TFNodes/yolo_evaluation_layer_1/strided_slice_48:0"] = 0x55555b614240, 
  ["yolo_evaluation_layer_1/concat_10:0_tx"] = 0x55555b613d20, 
  ["TFNodes/yolo_evaluation_layer_1/Reshape_17:0"] = 0x55555b6134b0, 
  ["TFNodes/yolo_evaluation_layer_1/mul_18:0"] = 0x55555b613240, 
  ["TFNodes/yolo_evaluation_layer_1/Sigmoid_7:0"] = 0x55555b613050, 
  ["TFNodes/yolo_evaluation_layer_1/Sigmoid_8:0"] = 0x55555b612840, 
  ["TFNodes/yolo_evaluation_layer_1/strided_slice_51:0"] = 0x55555b612220, 
  ["TFNodes/yolo_evaluation_layer_1/Reshape_15__87:0"] = 0x55555b611a60, 
  ["TFNodes/yolo_evaluation_layer_1/Reshape_15/shape_Concat__36:0"] = 0x55555b6114b0, 
  ["TFNodes/yolo_evaluation_layer_1/concat_6:0"] = 0x55555b610ea0, 
  ["TFNodes/yolo_evaluation_layer_1/Tile_5/multiples_Concat__46:0"] = 0x55555b6105d0, 
  ["TFNodes/yolo_evaluation_layer_1/Reshape_15/shape_Unsqueeze__32:0"] = 0x55555b60da10, 
  ["TFNodes/yolo_evaluation_layer_1/arange_4__77_loop:1"] = 0x55555b60e520, 
  ["TFNodes/yolo_evaluation_layer_1/arange_4__77_loop:0"] = 0x55555b60e450, 
  ["TFNodes/yolo_evaluation_layer_1/arange_4__77_trip_cnt:0"] = 0x55555b60d780, 
  ["TFNodes/yolo_evaluation_layer_1/arange_4__77_div:0"] = 0x55555b60cfc0, 
  ["TFNodes/yolo_evaluation_layer_1/arange_4__77_cast_diff:0"] = 0x55555b60cd40, 
  ["TFNodes/yolo_evaluation_layer_1/arange_4__77_diff__78:0"] = 0x55555b60ca50, 
  ["TFNodes/yolo_evaluation_layer_1/strided_slice_44:0"] = 0x55555b5cb8a0, 
  ["TFNodes/yolo_evaluation_layer_1/Reshape_14:0"] = 0x55555b5c8b10, 
  ["TFNodes/yolo_evaluation_layer_1/arange_5__52_trip_cnt:0"] = 0x55555b5caa10, 
  ["TFNodes/yolo_evaluation_layer_1/arange_5__52_ceil:0"] = 0x55555b5ca640, 
  ["TFNodes/yolo_evaluation_layer_1/arange_5__52_cast_diff:0"] = 0x55555b5c9e90, 
  ["TFNodes/yolo_evaluation_layer_1/arange_5__52_diff__53:0"] = 0x55555b5c9b50, 
  ["TFNodes/yolo_evaluation_layer_1/strided_slice_45:0"] = 0x55555b604060, 
  ["TFNodes/yolo_evaluation_layer_1/Reshape_3__306:0"] = 0x55555b5d8770, 
  ["TFNodes/yolo_evaluation_layer_1/Reshape_15/shape_Unsqueeze__33:0"] = 0x55555b5c8930, 
  ["TFNodes/yolo_evaluation_layer_1/strided_slice_13:0"] = 0x55555b5da1c0, 
  ["TFNodes/yolo_evaluation_layer_1/Cast_8:0"] = 0x55555b611150, 
  ["batch_norm_output_buffer64"] = 0x55555b59c670, ["model_1/add_23/add:0"] = 0x55555b5c44e0, 
  ["model_1/leaky_re_lu_52/LeakyRelu:0"] = 0x55555b5c42c0, 
  ["TFNodes/yolo_evaluation_layer_1/mul_7:0"] = 0x55555b5f7e80, 
  ["convolution_output23"] = 0x55555b5c3a10, ["convolution_output57"] = 0x55555b5a4ed0, 
  ["TFNodes/yolo_evaluation_layer_1/strided_slice_1:0"] = 0x55555b5cd130, 
  ["model_1/leaky_re_lu_51/LeakyRelu:0"] = 0x55555b5c3630, 
  ["TFNodes/yolo_evaluation_layer_1/Shape_2:0"] = 0x55555b5e9b70, 
  ["TFNodes/yolo_evaluation_layer_1/strided_slice_7__219:0"] = 0x55555b5ce4a0, 
  ["model_1/leaky_re_lu_68/LeakyRelu:0"] = 0x55555b5fec20, ["convolution_output25"] = 0x55555b5c1e80, 
  ["model_1/leaky_re_lu_49/LeakyRelu:0"] = 0x55555b5c1aa0, 
  ["TFNodes/yolo_evaluation_layer_1/Sigmoid_5:0"] = 0x55555b5f65f0, 
  ["model_1/add_21/add:0"] = 0x55555b5c0dc0, ["model_1/leaky_re_lu_48/LeakyRelu:0"] = 0x55555b5c0ba0, 
  ["convolution_output12"] = 0x55555b5e6930, ["batch_norm_output_buffer25"] = 0x55555b5bfb90, 
  ["model_1/add_20/add:0"] = 0x55555b5bf220, ["model_1/leaky_re_lu_45/LeakyRelu:0"] = 0x55555b5be3d0, 
  ["convolution_output30"] = 0x55555b5bdb20, ["model_1/add_6/add:0"] = 0x55555b5a3e10, 
  ["model_1/leaky_re_lu_44/LeakyRelu:0"] = 0x55555b5bd7d0, 
  ["batch_norm_output_buffer40"] = 0x55555b5b2e00, ["batch_norm_output_buffer28"] = 0x55555b5bd450, 
  ["convolution_output34"] = 0x55555b5ba3a0, 
  ["TFNodes/yolo_evaluation_layer_1/Reshape_9:0"] = 0x55555b5f5b40, 
  ["convolution_output54"] = 0x55555b5a79a0, 
  ["TFNodes/yolo_evaluation_layer_1/concat_10:0"] = 0x55555b613800, 
  ["model_1/leaky_re_lu_43/LeakyRelu:0"] = 0x55555b5bc7e0, 
  ["TFNodes/yolo_evaluation_layer_1/strided_slice_37:0"] = 0x55555b5fc2c0, 
  ["model_1/leaky_re_lu_50/LeakyRelu:0"] = 0x55555b5c2730, 
  ["model_1/leaky_re_lu_41/LeakyRelu:0"] = 0x55555b5bac50, 
  ["batch_norm_output_buffer48"] = 0x55555b5aa6b0, ["batch_norm_output_buffer32"] = 0x55555b5b9c40, 
  ["convolution_output35"] = 0x55555b5b9750, ["convolution_output28"] = 0x55555b5bf6a0, 
  ["TFNodes/yolo_evaluation_layer_1/arange__271_diff__272:0"] = 0x55555b5d3900, 
  ["TFNodes/yolo_evaluation_layer_1/Shape_3:0"] = 0x55555b602a50, 
  ["model_1/add_17/add:0"] = 0x55555b5b92e0, 
  ["TFNodes/yolo_evaluation_layer_1/arange_5__52_loop:1"] = 0x55555b609270, 
  ["model_1/leaky_re_lu_5/LeakyRelu:0"] = 0x55555b599fb0, 
  ["batch_norm_output_buffer15"] = 0x55555b5c7f60, 
  ["model_1/leaky_re_lu_32/LeakyRelu:0"] = 0x55555b5b3180, ["model_1/add_19/add:0"] = 0x55555b5bca00, 
  ["batch_norm_output_buffer34"] = 0x55555b5b80b0, ["model_1/add_16/add:0"] = 0x55555b5b7750, 
  ["convolution_output38"] = 0x55555b5b6c80, 
  ["TFNodes/yolo_evaluation_layer_1/mul_16:0"] = 0x55555b615660, 
  ["convolution_output18"] = 0x55555b5c7a30, 
  ["TFNodes/yolo_evaluation_layer_1/Squeeze:0"] = 0x55555b595170, 
  ["convolution_output39"] = 0x55555b5b6030, ["convolution_output10"] = 0x55555b5fd750, 
  ["TFNodes/yolo_evaluation_layer_1/Tile:0"] = 0x55555b5d48c0, 
  ["batch_norm_output_buffer16"] = 0x55555b5c7330, ["batch_norm_output_buffer37"] = 0x55555b5b5620, 
  ["TFNodes/yolo_evaluation_layer_1/mul_17:0"] = 0x55555b619750, 
  ["batch_norm_output_buffer38"] = 0x55555b5b4990...}

@daquexian could you provide any guidance on how to fix this?

aitikgupta commented 4 years ago

Any updates on this?

LewsTherin511 commented 3 years ago

Having the same error while trying to convert a custom SSD object detection model. The original model was implemented in GluonCV, fine-tuned from one of the models in the ZOO, and converted to ONNX format. Any updates?