pytorch -> ONNX -> Tensorflow -> TensorflowLite

[sahasraa]

Hi all,

1. I am trying to run yolov7.pt model on raspberry. as pytorch doesn't directly have supporting packages to run on raspi, I am following this pytorch -> ONNX -> Tensorflow -> TensorflowLite.
2. I am working on MacBook AIR M2
3. I have converted pytorch to ONNX successfully. exporting ONNX to TF has been a struggle

The issues I am facing are, package version issues.

apple silicon supports TensorFlow-mascos package, that is having package dependency issues with others.

The command I am using is onnx-tf convert -i yolov7.onnx -o yolov7.pb

The error I am getting is,

ValueError: Arg specs do not match: original=FullArgSpec(args=['input', 'dtype', 'name', 'layout'], varargs=None, varkw=None, defaults=(None, None, None), kwonlyargs=[], kwonlydefaults=None, annotations={}), new=FullArgSpec(args=['input', 'dtype', 'name'], varargs=None, varkw=None, defaults=(None, None), kwonlyargs=[], kwonlydefaults=None, annotations={}), fn=<function ones_like_v2 at 0x16c595ca0>

This means result of version mismatch between TensorFlow and TensorFlow Probability. The core issue here is that the arguments to a function in TensorFlow (ones_like_v2) don't match the expectations in TensorFlow Probability.

Could anyone help to help solve of mismatch between versions.

Thanks in advance