errors broke out during compliation of tflite file for edgetpu

[jisooyu]

Instructions followed: compilation instructions listed on https://coral.withgoogle.com/docs/edgetpu/retrain-detection/ OS: Host: Mac Mojavee 10.14.8 Local: Ubunbu 18.04 on Virtual Box Scripts ran : jisooyu@jisooyu-VirtualBox:/usr/bin/edgetpu_compiler_bin$ edgetpu_compiler -s output_tflite_graph.tflite Output generated: Edge TPU Compiler version 2.0.267685300

Model compiled successfully in 313 ms.

Input model: output_tflite_graph.tflite Input size: 5.33MiB Output model: output_tflite_graph_edgetpu.tflite Output size: 5.75MiB On-chip memory available for caching model parameters: 7.62MiB On-chip memory used for caching model parameters: 5.66MiB Off-chip memory used for streaming uncached model parameters: 0.00B Number of Edge TPU subgraphs: 1 Total number of operations: 64 Operation log: output_tflite_graph_edgetpu.log

Model successfully compiled but not all operations are supported by the Edge TPU. A percentage of the model will instead run on the CPU, which is slower. If possible, consider updating your model to use only operations supported by the Edge TPU. For details, visit g.co/coral/model-reqs. Number of operations that will run on Edge TPU: 63 Number of operations that will run on CPU: 1

Operator Count Status

CUSTOM 1 Operation is working on an unsupported data type CONCATENATION 2 Mapped to Edge TPU CONV_2D 34 Mapped to Edge TPU DEPTHWISE_CONV_2D 13 Mapped to Edge TPU RESHAPE 13 Mapped to Edge TPU LOGISTIC 1 Mapped to Edge TPU Error opening file for writing: output_tflite_graph_edgetpu.tflite

Anyone please help me to find what is wrong. The error doesn't tell much.

[zvit3k]

Hi @jisooyu

As you can see model is correctly compiled. The issue is probably related to the privileges on the command you want to execute.

Try sudo edgetpu_compiler output_tflite_graph.tflite

I hope it helps. Witek

[mbcel]

Furthermore it seems that as the compilation states that your whole model will not completely run on the edgetpu which probably results in very slow inference time. Dpending on the possition of the unsupported layer.

[Ryan-Johnson-1315]

I am having the same problem. The model will compile successfully. When I run it on with my TPU it throws these errors: Null custom op data Node number 0 (edgetpu-custom-op) failed to prepare.

Any suggestions?

[Namburger]

@jisooyu You most likely just don't have permission to write in that directory. Moving the model to another directory and compile should fix the problem, or just give yourself permission for the directory with:

sudo chown -R $USER:$USER dir_name

The problem with compiling with sudo is that it'll most likely write the model as root and then your regular user may not get permission to use the model haha

@Ryan-Johnson-1315 hard to tell without more details but that's most likely because you didn't register the edgetpu ops. Python would looks like this:

interpreter = tflite.Interpreter(model_path,
  experimental_delegates=[tflite.load_delegate('libedgetpu.so.1')])

C++ would look like this:

  tflite::ops::builtin::BuiltinOpResolver resolver;
  resolver.AddCustom(edgetpu::kCustomOp, edgetpu::RegisterCustomOp());