分段错误（核心转储） - TFLite

Question

Describe the problem

To read a model from official TensorFlow source (COCO SSD MobileNet v1) and perform inference with minimal.cc , we get the error below.

System information

• Host OS Platform and Distribution: Linux Ubuntu 16.04
• TensorFlow installed from (source or binary): From source (branch r1.12)
• Target platform: iMX.6 (Arm v7)

Please provide the exact sequence of commands/steps when you ran into the problem

root@analytics:~# ./minimal ssd_mobilenet_v1_1_metadata_1.tflite
minimal: /usr/src/debug/tensorflow-lite/1.0-r0/git/tensorflow/contrib/lite/tools/make/downloads/eigen/unsupported/Eigen/CXX11/src/util/
MaxSizeVector.h:84: T& EigenForTFLite::MaxSizeVector<T>::operator[](std::size_t) [with T = EigenForTFLite::RunQueue
<EigenForTFLite::StlThreadEnvironment::Task, 1024u>*; std::size_t = unsigned int]: 
Assertion `i < size_' failed. 
Segmentation fault (core dumped)

Do you have any idea or suggestion about why we are facing this error (because of versions, TFlite binaries or the model etc.)? Thank you in advance.

Answer 1

Looks like the TensorFlow Lite version is too old to be supported. Please consider using TF 2.5 or beyonds.

Answer 2

hello，you should care tflite where come from, it is error that tflite version

Answer 3

I faced something similar, my interpreter object was giving Segmentation fault (core dumped) error. On debugging I found out that the issue was in

• .so file of tflite_runtime, got corrupted while uploading on the server.
• model files, got corrupted because of same reason.

Then I used git to transfer files between servers, it helped.