A/libc: Fatal signal 11 (SIGSEGV), code 1, fault addr 0x70 in tid 28007

[kazuimotn]

There was a problem when calling converted custom model (LSTM.tflite) in Android Studio.
with

converter.experimental_new_converter = True

System information
model training machine

• Ubuntu 16.04 (LTS)
• GTX1080
• tensorflow 1.13.1
• Keras 2.2.4
• Python 3.6.8

model converting & loading machine

• MacOS Mojave 10.14.5
• Python 3.7.1
• tensorflow 2.0.0, tf-nightly 2.1.0.dev20191113
• Android Studio 3.4.1
• targetSdkVersion 28

Smartphone

• Android 8.0.0
• SONY SO-04J (docomo xperia)

I can convert custom LSTM model from .h5 to .tflite with following code:

import tensorflow as tf
model = tf.keras.models.load_model("LSTM.h5")
converter = tf.lite.TFLiteConverter.from_keras_model(model)
converter.experimental_new_converter = True
tflite_model = converter.convert()
open("LSTM.tflite", "wb").write(tflite_model)

Next I want to load this model (LSTM.tflite) into my Android APK.
I already have class that can perform inference using 1D-CNN custom model (1DCNN.tflite).

package iis.kmjlab.kazuimotn.sartips.others;
import android.content.res.AssetFileDescriptor;
import android.content.res.AssetManager;
import org.tensorflow.lite.Interpreter;
import java.io.FileInputStream;
import java.io.IOException;
import java.nio.ByteBuffer;
import java.nio.MappedByteBuffer;
import java.nio.channels.FileChannel;

public class TensorFlowLiteClassifier {

    private Interpreter interpreter;
    public static final String MODEL_FILE = "1DCNN.tflite";
    public static final int LABEL_NUM = 2;

    /**
     * Registers interpreter
     */
    private TensorFlowLiteClassifier(Interpreter interpreter) {
        this.interpreter = interpreter;
    }

    /**
     * Loads model into interpreter
     */
    public static TensorFlowLiteClassifier classifier(AssetManager assetManager, String modelPath) throws IOException {
        ByteBuffer byteBuffer = loadModelFile(assetManager, modelPath);
        Interpreter interpreter = new Interpreter(byteBuffer);
        return new TensorFlowLiteClassifier(interpreter);
    }

    /**
     * Loads model
     */
    private static MappedByteBuffer loadModelFile(AssetManager assets, String path) throws IOException {
        AssetFileDescriptor file = assets.openFd(path);
        FileInputStream stream = new FileInputStream(file.getFileDescriptor());
        FileChannel channel = stream.getChannel();
        long startOffset = file.getStartOffset();
        long declaredLength = file.getDeclaredLength();
        return channel.map(FileChannel.MapMode.READ_ONLY, startOffset, declaredLength);
    }

    /**
     * Function that actually performs inference
     */
    public float[][] predictProbabilities(float[][] input) {
        float[][] output = new float[1][LABEL_NUM];
        interpreter.run(input, output);
        return output;
    }
}

When I use 1DCNN.tflite, this class works completely. However, when I change the model into LSTM.tflite, getting ERROR at this line

Interpreter interpreter = new Interpreter(byteBuffer);

And ERROR logcat is following
A/libc: Fatal signal 11 (SIGSEGV), code 1, fault addr 0x70 in tid 28007

How can I deal with this unknown error?
After my some investigation, this error seems to occur when the prepared array is exceeded for some reason.

[haozha111]

I'm wondering if it's an issue with the Android APK environment. Does the same lstm flatbuffer model run on desktop?

[kazuimotn]

I can infer probabilities using same LSTM.tflite model with this Python code:

import tensorflow as tf
interpreter = tf.compat.v2.lite.Interpreter("LSTM.tflite")
interpreter.allocate_tensors()
input  = interpreter.tensor(interpreter.get_input_details()[0]["index"])
output = interpreter.tensor(interpreter.get_output_details()[0]["index"])
for i in range(10):
  input().fill(1)
  interpreter.invoke()
  print("inference %s" % output())

Output is following:

inference [[9.5466515e-09 1.0000000e+00]]
inference [[9.5466515e-09 1.0000000e+00]]
inference [[9.5466515e-09 1.0000000e+00]]
inference [[9.5466515e-09 1.0000000e+00]]
inference [[9.5466515e-09 1.0000000e+00]]
inference [[9.5466515e-09 1.0000000e+00]]
inference [[9.5466515e-09 1.0000000e+00]]
inference [[9.5466515e-09 1.0000000e+00]]
inference [[9.5466515e-09 1.0000000e+00]]
inference [[9.5466515e-09 1.0000000e+00]]

My model is for binary classification, it means this output is what I want.
In other words, does it mean that the Android APK environment is incorrect?

[kazuimotn]

It might be due to some native methods here:
TensorflowLite/library/src/main/java/org/tensorflow/lite/NativeInterpreterWrapper.java

private static native long createErrorReporter(int size);
private static native long createModelWithBuffer(MappedByteBuffer modelBuffer, long errorHandle);
private static native long createInterpreter(long modelHandle);

/**
   * Initializes a {@code NativeInterpreterWrapper} with a {@code MappedByteBuffer}. The
   * MappedByteBuffer should not be modified after the construction of a {@code
   * NativeInterpreterWrapper}.
   */
  NativeInterpreterWrapper(MappedByteBuffer mappedByteBuffer) {
    modelByteBuffer = mappedByteBuffer;
    errorHandle = createErrorReporter(ERROR_BUFFER_SIZE);
    modelHandle = createModelWithBuffer(modelByteBuffer, errorHandle);
    interpreterHandle = createInterpreter(modelHandle);
  }

[lucasveljacic]

Hi!, could any of you solve this issue?
I'm having the same error with an LSTM clasifier.
In fact, not exactly the same... actually I'm getting this:

java.lang.IllegalArgumentException: ByteBuffer is not a valid flatbuffer model

[luang008]

I solved a similar issue following this guide.

Particularly adding this into build.gradle:

android { defaultConfig { ndk { abiFilters 'armeabi-v7a', 'arm64-v8a' } } }

to filter unnecessary ABI dependencies.

[Saduf2019]

@kazuimotn
Please update as per above comment.

[google-ml-butler]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you.

[google-ml-butler]

Closing as stale. Please reopen if you'd like to work on this further.

[google-ml-butler]

Are you satisfied with the resolution of your issue?
Yes
No

[StewardH]

I am seeing the same crash: Fatal signal 11 (SIGSEGV), code 1 (SEGV_MAPERR)

This occurs while running the standard object detection routine using the mobileNet SSD .tflite model that can be found in tf model zoo:

The issue was replicated on Samsung Galaxy S20+ as well as on Motorola phones and it seems like this crash is phone independent. On Motorola phones the crash occurs within 2 hours while running the detection while on Samsung Galaxy the issue sometimes occurs after 8h of running the same code on the same input data in a repetitive loop. Filtering the abi's did not affect this case.
Did anyone resolved this crash in a more general fashion?

It seems like it occurs on the RenderThread during the DeferredLayerUpdate.
Here is my crash report:
--------- beginning of crash
2022-02-13 15:33:42.474 18219-18249/? A/libc: Fatal signal 11 (SIGSEGV), code 1 (SEGV_MAPERR), fault addr 0x2820 in tid 18249 (RenderThread), pid 18219 (dia.rec)
2022-02-13 15:33:42.626 18717-18717/? A/DEBUG: *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** ***
2022-02-13 15:33:42.626 18717-18717/? A/DEBUG: Build fingerprint: 'motorola/kyoto_retaile/kyoto:11/RRKS31.Q3-19-97-3/bc988:user/release-keys'
2022-02-13 15:33:42.626 18717-18717/? A/DEBUG: Revision: 'PVT1'
2022-02-13 15:33:42.626 18717-18717/? A/DEBUG: ABI: 'arm64'
2022-02-13 15:33:42.628 18717-18717/? A/DEBUG: Timestamp: 2022-02-13 15:33:42-0500
2022-02-13 15:33:42.628 18717-18717/? A/DEBUG: pid: 18219, tid: 18249, name: RenderThread >>> dia.rec <<<
2022-02-13 15:33:42.628 18717-18717/? A/DEBUG: uid: 10324
2022-02-13 15:33:42.628 18717-18717/? A/DEBUG: signal 11 (SIGSEGV), code 1 (SEGV_MAPERR), fault addr 0x2820
2022-02-13 15:33:42.628 18717-18717/? A/DEBUG: x0 b4000076636106d0 x1 b4000076636dbed0 x2 0000007576fea8c0 x3 00000000000000ff
2022-02-13 15:33:42.628 18717-18717/? A/DEBUG: x4 0000000000000003 x5 b4000076c35c8980 x6 b4000076c35c8980 x7 0000000000000000
2022-02-13 15:33:42.628 18717-18717/? A/DEBUG: x8 0000000000000000 x9 0000000000000001 x10 0000000000000000 x11 0000000063610768
2022-02-13 15:33:42.628 18717-18717/? A/DEBUG: x12 00000000e1b5b851 x13 2365000000084108 x14 001c222856e7e2a0 x15 000021d5f589e147
2022-02-13 15:33:42.628 18717-18717/? A/DEBUG: x16 00000078764327f8 x17 00000078739dbb70 x18 0000007576d54000 x19 b4000076636106d0
2022-02-13 15:33:42.628 18717-18717/? A/DEBUG: x20 b4000076636dbed0 x21 b4000076636dbef0 x22 0000000000002800 x23 b4000075f35b75b8
2022-02-13 15:33:42.628 18717-18717/? A/DEBUG: x24 b4000076235d4c50 x25 0000007576feacc0 x26 0000007576feaff8 x27 00000000000fe000
2022-02-13 15:33:42.628 18717-18717/? A/DEBUG: x28 00000000000fc000 x29 0000007576fea870
2022-02-13 15:33:42.628 18717-18717/? A/DEBUG: lr 0000007874205500 sp 0000007576fea870 pc 000000787420ae00 pst 0000000020001000
2022-02-13 15:33:42.669 18717-18717/? A/DEBUG: backtrace:
2022-02-13 15:33:42.669 18717-18717/? A/DEBUG: #00 pc 0000000000239e00 /system/lib64/libhwui.so (android::uirenderer::Layer::Layer(android::uirenderer::RenderState&, sk_sp, int, SkBlendMode)+304) (BuildId: 0e3f217960712f3e10c4b8b3b618903d)
2022-02-13 15:33:42.669 18717-18717/? A/DEBUG: #1 pc 00000000002344fc /system/lib64/libhwui.so (android::uirenderer::DeferredLayerUpdater::apply()+428) (BuildId: 0e3f217960712f3e10c4b8b3b618903d)
2022-02-13 15:33:42.669 18717-18717/? A/DEBUG: #2 pc 000000000021fbd8 /system/lib64/libhwui.so (_ZNSt3__110__function6__funcIZN7android10uirenderer12renderthread13DrawFrameTask11postAndWaitEvE3$_0NS_9allocatorIS6_EEFvvEEclEv$c303f2d2360db58ed70a2d0ac7ed911b+684) (BuildId: 0e3f217960712f3e10c4b8b3b618903d)
2022-02-13 15:33:42.669 18717-18717/? A/DEBUG: #3 pc 000000000020e504 /system/lib64/libhwui.so (android::uirenderer::WorkQueue::process()+220) (BuildId: 0e3f217960712f3e10c4b8b3b618903d)
2022-02-13 15:33:42.669 18717-18717/? A/DEBUG: #4 pc 000000000022f41c /system/lib64/libhwui.so (android::uirenderer::renderthread::RenderThread::threadLoop()+88) (BuildId: 0e3f217960712f3e10c4b8b3b618903d)
2022-02-13 15:33:42.669 18717-18717/? A/DEBUG: #5 pc 0000000000015414 /system/lib64/libutils.so (android::Thread::_threadLoop(void*)+260) (BuildId: 82f928b900a93dc07b75aefd76a59775)
2022-02-13 15:33:42.669 18717-18717/? A/DEBUG: #6 pc 0000000000014cd8 /system/lib64/libutils.so (thread_data_t::trampoline(thread_data_t const*)+412) (BuildId: 82f928b900a93dc07b75aefd76a59775)
2022-02-13 15:33:42.669 18717-18717/? A/DEBUG: #7 pc 00000000000af86c /apex/com.android.runtime/lib64/bionic/libc.so (__pthread_start(void*)+64) (BuildId: ced3426ba03689dee1bf74b8de9c436a)
2022-02-13 15:33:42.669 18717-18717/? A/DEBUG: #8 pc 0000000000050110 /apex/com.android.runtime/lib64/bionic/libc.so (__start_thread+64) (BuildId: ced3426ba03689dee1bf74b8de9c436a)

[yen-dang-ts]

I got same issue