Back to DrivenData | Blog

Issue adding onnxruntime

Hello and happy holidays,

I am having an issue adding the onnxruntime module as a dependency. I am a Matlab user and in order to use my models in Python, I export them from Matlab as onnx files and use them in Python with the onnxruntime module.

I followed the add dependencies instructions, but I get the following error:

Check failure on line 1178 in Build log

@azure-pipelinesazure-pipelines/ drivendataorg.ai4earth-serengeti-runtime (test_cpu_installs)

Build log #L1178

Bash exited with code '125'.

I added -onnxruntime=1.1.0 to py-cpu.yml
I added onnxruntime under ## ADD ADDITIONAL REQUIREMENTS BELOW HERE ## in test-installs.py

If anyone can make a suggestion or add the dependency that will would be great.

Regards,

Hello,

I resolved my problem by converting my Onnx model to a Tensorflow model using onnx_tf. The resulting pb file was used in my main.py file. This method keeps me from having to add anything new to the submission environment. Just note, if your Onnx model has a ceil function in a layer, you will need to use a Tensorflow version less than 2.0 for the conversion.

1 Like