TypeError: injectFaultSlice() takes exactly 1 argument (3 given)

[abhishektyaagi]

Hi,
I am trying to run TensorFI on an object detection framework. I am using TensorFlow 1.15 and the algorithm of Object Detection is implemented in Python 2.7 and tested with Tensorflow 1.2 and 1.4.

I just initialise the injection phase by using
fi = ti.TensorFI(sess, name = "FrustumPointNet", logLevel = 50,disableInjections = False)

And that is the only change I do the original code which runs fine. The error that I am getting is:

Log:
2021-06-11 16:29:40.268495: W tensorflow/core/framework/op_kernel.cc:1639] Invalid argument: exceptions.TypeError: injectFaultSlice() takes exactly 1 argument (3 given) Traceback (most recent call last): File "/u/atyagi2/anaconda3/envs/tf1.15/lib/python2.7/site-packages/tensorflow_core/python/ops/script_ops.py", line 235, in __call__ ret = func(*args) TypeError: injectFaultSlice() takes exactly 1 argument (3 given)

This is my first time using the tool and hence I am not able to trace back the origin of this error. I would appreciate if you can point out what might be going wrong here.

P.S: I am able to run the DNN-Modles examples given, so I don't know where things are going wrong.

[nniranjhana]

Hi @abhishek-t-naive thanks for your question, could you change the logLevel to 10 or 100 (I forget which gives more information) so that we can read more of the error log? You can attach the full log as a file (it might be big).. from what I can see, it looks like somewhere in the middle an injection function is getting more arguments than it's supposed to.

Could you also attach the example/model code for which you're trying to inject into (the object detection Python code) so I could see what operators it uses and if TensorFI supports all those operators.

Oh and also the YAML config file you're using. Thanks