command hailomz compile doesn't work with error ONNX shape inference failed, please take a look to my github issue

General
, ,
1

Please take a look github issue,
I described it with screenshots.

I run it on my macbook but in docker of ubuntu. Sshould be fine. take a look at my dockerfile in github issue

I tried to install 3.30.0 and 2.14.0 versions and it was lib issue with numpy.

When I tried to install 3.28.0 and 2.12.0 versions and it installed successfully:

hailo_dataflow_compiler-3.28.0-py3-none-linux_x86_64.whl
hailo_model_zoo-2.12.0-py3-none-any.whl

but still I got error after running this command:

hailomz compile yolov8s --ckpt=best.onnx --hw-arch hailo8l --calib-path train/images --classes 3 --performance

ONNX shape inference failed: [ONNXRuntimeError] : 2 : INVALID_ARGUMENT : Failed to load model with error: /onnxruntime_src/onnxruntime/core/graph/model.cc:147 onnxruntime::Model::Model(onnx::ModelProto&&, const PathString&, const IOnnxRuntimeOpSchemaRegistryList*, const onnxruntime::logging::Logger&, const onnxruntime::ModelOptions&) Unsupported model IR version: 9, max supported IR version: 8

[info] Unable to simplify the model: '/model.2/Split_output_0

also I tried 3.29.0 2.13.0 versions like: hailo_dataflow_compiler-3.29.0-py3-none-linux_x86_64.whl and hailo_model_zoo-2.13.0-py3-none-any.whl and it still doesn’t work because of numpy version incorrection.

The question why my command doesn’t work?

hailomz compile yolov8s --ckpt=best.onnx --hw-arch hailo8l --calib-path train/images --classes 3 --performance
with these successfully installed libs?

hailo_dataflow_compiler-3.28.0-py3-none-linux_x86_64.whl
hailo_model_zoo-2.12.0-py3-none-any.whl
1 Like
2

I got same issue. Do you have any updated?

3

I also try use docker, install DFC 3.30.0 and hailomz 1.12 and got issue when Quantization-Aware Fine-Tuning “hailo_model_optimization.acceleras.utils.acceleras_exceptions.SubprocessUnexpectedFailure: Subprocess step2 failed with unexpected error. exitcode -7”