ERROR: Failed building wheel for onnx

[meremeev]

Building converter for TorchScript backend. Got error below

/tmp/pip-install-92s0sdmt/onnx_38ddb31565e7418686210a037f41ec09/onnx/defs/sequence/defs.cc: In function ‘bool onnx::BuildSequenceMapBodyFunc(const onnx::FunctionBodyBuildContext&, const onnx::OpSchema&, onnx::FunctionProto&)’:
  /tmp/pip-install-92s0sdmt/onnx_38ddb31565e7418686210a037f41ec09/onnx/defs/sequence/defs.cc:675:40: error: no match for ‘operator[]’ (operand types are ‘google::protobuf::RepeatedPtrField<onnx::ValueInfoProto>’ and ‘int’)
           seq_at_node.add_output(g_inputs[inputIndex].name());
                                          ^
  /tmp/pip-install-92s0sdmt/onnx_38ddb31565e7418686210a037f41ec09/onnx/defs/sequence/defs.cc:683:37: error: no match for ‘operator[]’ (operand types are ‘google::protobuf::RepeatedPtrField<onnx::ValueInfoProto>’ and ‘int’)
           identity.add_output(g_inputs[inputIndex].name());
                                       ^
  /tmp/pip-install-92s0sdmt/onnx_38ddb31565e7418686210a037f41ec09/onnx/defs/sequence/defs.cc:741:30: error: no match for ‘operator[]’ (operand types are ‘google::protobuf::RepeatedPtrField<onnx::ValueInfoProto>’ and ‘int’)
       int64_t dtype = g_outputs[outputIndex].type().tensor_type().elem_type();

The same error if I try to use pre-compiled binaries - mmdeploy-0.5.0-linux-x86_64-cuda10.2-tensorrt8.2.3.0

Also

mark@MEREMEEV-LT:~/Code/mmdeploy$ python3 tools/check_env.py
Segmentation fault (core dumped)

[RunningLeon]

@meremeev Hi, sorry for the late reply. It's a bit strange. Could you share the full script for building mmdeploy? BTW, could you check where leads to the segmentation fault with tools/check_env.py? If possible, comment

mmdeploy/tools/check_env.py

Line 30 in 0e1a3aa

import mmdeploy.apis.tensorrt as trt_apis

and try again.

[Talnex]

HiHi,
I also had the same problem, but I found a solution
This is because your Protobuf's version is too low, the official recommendation is above 3.12.2
or in the latest ONNX version this problem is solved https://github.com/onnx/onnx/pull/4354
you can clone and build ONNX from source ;)