I am using a part of SNPE 1.13.0 android example to execute mobilenet-ssd network.
I have a 640x480 YUV frame that i convert to 300x300 RGB bitmap and then feed to mobilenet-ssd. While doing execute() on 1st frame itself, I get following error:
04-17 16:27:53.790 3138 3211 E AndroidRuntime: FATAL EXCEPTION: player_main
04-17 16:27:53.790 3138 3211 E AndroidRuntime: Process: veg.mediaplayer.sdk.test.view2x2, PID: 3138
04-17 16:27:53.790 3138 3211 E AndroidRuntime: java.lang.IllegalStateException: Failed to forward propagate inputs! Cause: error_code=809; error_message=OpenCL function has returned error. OpenCL Error (-7) CL_PROFILING_INFO_NOT_AVAILABLE; error_component=GPU Runtime; line_no=233; thread_id=-703366864; opencl_error=-7
04-17 16:27:53.790 3138 3211 E AndroidRuntime: at com.qualcomm.qti.snpe.internal.NativeNetwork.nativeForward(Native Method)
04-17 16:27:53.790 3138 3211 E AndroidRuntime: at com.qualcomm.qti.snpe.internal.NativeNetwork.execute(NativeNetwork.java:120)
04-17 16:27:53.790 3138 3211 E AndroidRuntime: at com.qualcomm.qti.snpe.imageclassifiers.tasks.ClassifyImageTask.classify(ClassifyImageTask.java:111)
04-17 16:27:53.790 3138 3211 E AndroidRuntime: at xxx.mediaplayer.sdk.test.view2x2.MainActivity.OnVideoRendererFrameAvailable(MainActivity.java:838)
04-17 16:27:53.790 3138 3211 E AndroidRuntime: at xxx.mediaplayer.sdk.MediaPlayer.OnVideoRendererFrameAvailable(MediaPlayer.java:2721)
04-17 16:27:53.792 974 1628 W ActivityManager: Force finishing activity xxx.mediaplayer.sdk.test.view2x2/.MainActivity
I have made sure that I have RGB floating point values ordered properly by running some example YUV to RGB sequences (e.g. https://www.mikekohn.net/file_formats/yuv_rgb_converter.php).
What's the reason for "Failed to forward propagate inputs!"?
public String[] classify(Bitmap image) { mResult.clear(); image.getPixels(pixels, 0, image.getWidth(), 0, 0, image.getWidth(), image.getHeight()); for (int y = 0; y < image.getHeight(); y++) { //Log.e("Test", "\n[" + y + "]"); for (int x = 0; x < image.getWidth(); x++) { final int rgb = pixels[y * image.getWidth() + x]; float[] pixelFloats = {Color.blue(rgb), Color.green(rgb), Color.red(rgb)}; mTensor.write(pixelFloats, 0, pixelFloats.length, y, x); } } final Map<String, FloatTensor> inputs = new HashMap<>(); inputs.put("Preprocessor/sub:0", mTensor); final Map<String, FloatTensor> outputs = mNeuralNetwork.execute(inputs);
Hi, any suggestions on this issue?
any suggestions on this? I see this very consistently. Note that when things work, I am able to have GPU do inference. So this is intermittent error!
Bump bump, Qualcomm! Seeing this too - no idea what it means.