Hi,
I am working in ubuntu 18.04 with an LPR engine that can deal with many inference engines: tflite, edgetpu (coral ai), tensorrt, opencv, Hailo chip.
I am adding the libtensorflow.so to the above repertoire.
We are inferencing on different edge platforms. one of them is Compulab fitlet2, it has Celeron CPU that does not support AVX AVX2.
When I link my program to libtensorflow.so the program crash on CPUs that does not support avx2. (program runs well and do nice inference on CPU that have the avx2 , intel i3 for example).
I tried loading the library dynamically from the code by dlopen and mapping the routines with dlsym. I then get a crash with TF_LoadSessionFromSavedModel:
2022-03-17 09:41:14.346300: I tensorflow/cc/saved_model/reader.cc:43] Reading SavedModel from: ./SavedModel
2022-03-17 09:41:14.384910: I tensorflow/cc/saved_model/reader.cc:107] Reading meta graph with tags { serve }
2022-03-17 09:41:14.384951: I tensorflow/cc/saved_model/reader.cc:148] Reading SavedModel debug info (if present) from: ./SavedModel
2022-03-17 09:41:14.385028: I tensorflow/core/platform/cpu_feature_guard.cc:151] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations: AVX2 FMA
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
2022-03-17 09:41:14.520482: I tensorflow/cc/saved_model/loader.cc:210] Restoring SavedModel bundle.
free(): invalid pointer
Aborted (core dumped)
Thanks
Shay
You need to recompile from sources or find a third party precompiled version: