TPU delegate erroring with seg fault

[stakach]

coral ai accelerator not working

Invalid memory access (signal 11) at address 0x5594a7c70330
[0x5594a6f90796] *Exception::CallStack::print_backtrace:Nil +118 in ./processor
[0x5594a6f6d576] ~procProc(Int32, Pointer(LibC::SiginfoT), Pointer(Void), Nil) +310 in ./processor
[0x7efdcd63c460] ?? +139628537693280 in /lib/x86_64-linux-gnu/libc.so.6
[0x7efdcf65333c] ?? +139628571341628 in /home/steve/projects/edge-ai/bin/libtensorflowlite_c.so
[0x7efdcf656e41] ?? +139628571356737 in /home/steve/projects/edge-ai/bin/libtensorflowlite_c.so
[0x7efdcf657a8a] ?? +139628571359882 in /home/steve/projects/edge-ai/bin/libtensorflowlite_c.so
[0x7efdcfd676a1] ?? +139628578764449 in /lib/x86_64-linux-gnu/libedgetpu.so.1
[0x7efdcf658f49] ?? +139628571365193 in /home/steve/projects/edge-ai/bin/libtensorflowlite_c.so
[0x7efdcf6595c2] ?? +139628571366850 in /home/steve/projects/edge-ai/bin/libtensorflowlite_c.so
[0x7efdcf63c738] ?? +139628571248440 in /home/steve/projects/edge-ai/bin/libtensorflowlite_c.so
[0x7efdcf6205f7] ?? +139628571133431 in /home/steve/projects/edge-ai/bin/libtensorflowlite_c.so
[0x7efdcf61f1fd] TfLiteInterpreterCreate +82 in /home/steve/projects/edge-ai/bin/libtensorflowlite_c.so
[0x5594a719bdd7] *TensorflowLite::Interpreter#initialize<TensorflowLite::Model, TensorflowLite::InterpreterOptions>:LibTensorflowLite::Status +55 in ./processor
[0x5594a719bd86] *TensorflowLite::Interpreter::new<TensorflowLite::Model, TensorflowLite::InterpreterOptions>:TensorflowLite::Interpreter +86 in ./processor
[0x5594a7193348] *TensorflowLite::Client#initialize<Path, (TensorflowLite::EdgeTPU::Delegate | Nil), Nil, (Array(String) | Nil), &Proc(String, Nil)>:TensorflowLite::Interpreter +280 in ./processor
[0x5594a7193218] *TensorflowLite::Client::new<Path, (TensorflowLite::EdgeTPU::Delegate | Nil), Nil, (Array(String) | Nil), &Proc(String, Nil)>:TensorflowLite::Client +264 in ./processor
[0x5594a7193104] *TensorflowLite::Client::new:delegate:labels<Path, (TensorflowLite::EdgeTPU::Delegate | Nil), (Array(String) | Nil)>:TensorflowLite::Client +100 in ./processor
[0x5594a7150609] *TensorflowLite::Pipeline::Configuration::Model +1273 in ./processor
[0x5594a714fc0c] *TensorflowLite::Pipeline::Configuration::Model +156 in ./processor
[0x5594a7158d77] *TensorflowLite::Pipeline::Coordinator#initialize<String, TensorflowLite::Pipeline::Configuration::Pipeline>:Float32 +487 in ./processor
[0x5594a7158b82] *TensorflowLite::Pipeline::Coordinator::new<String, TensorflowLite::Pipeline::Configuration::Pipeline>:TensorflowLite::Pipeline::Coordinator +98 in ./processor
[0x5594a712d562] *EdgeAI::Processor#start_stream<TensorflowLite::Pipeline::Configuration::Pipeline>:(Fiber | Nil) +402 in ./processor
[0x5594a712d396] *EdgeAI::Processor#start_streams:Nil +326 in ./processor
[0x5594a6f55a60] __crystal_main +1952 in ./processor
[0x5594a7202cd6] *Crystal::main_user_code<Int32, Pointer(Pointer(UInt8))>:Nil +6 in ./processor
[0x5594a7202c4a] *Crystal::main<Int32, Pointer(Pointer(UInt8))>:Int32 +58 in ./processor
[0x5594a6f63536] main +6 in ./processor
[0x7efdcd623a90] ?? +139628537592464 in /lib/x86_64-linux-gnu/libc.so.6
[0x7efdcd623b49] __libc_start_main +137 in /lib/x86_64-linux-gnu/libc.so.6
[0x5594a6f551f5] _start +37 in ./processor
[0x0] ???

[stakach]

works with TensorFlow version

# clone the required repositories
git clone --depth 1 --branch "v2.6.0" https://github.com/tensorflow/tensorflow

also works with v2.8.4