Incubator-mxnet: Segmentation fault:11 with Python 2.7.11 OSX

[shadowinlife]

For bugs or installation issues, please provide the following information.
The more information you provide, the more likely people will be able to help you.

Environment info

Operating System:
OS X 10.11.6

Compiler:
Default Compile Turtoris

Package used (Python/R/Scala/Julia):
Python
MXNet version installed from source:
0.9.1
MXNet commit hash (git rev-parse HEAD):
5656c8601265a437c1cb7ea18a6f1661f346c8b5

Python version and distribution:
2.7.11

Error Message:

Please paste the full error message, including stack trace.

System Integrity Protection: enabled

Crashed Thread:        8

Exception Type:        EXC_BAD_ACCESS (SIGSEGV)
Exception Codes:       KERN_INVALID_ADDRESS at 0x000000047d746000

VM Regions Near 0x47d746000:
    VM_ALLOCATE            000000047b746000-000000047d746000 [ 32.0M] rw-/rwx SM=PRV  
--> VM_ALLOCATE            000000047d746000-000000047d747000 [    4K] rw-/rwx SM=ALI  
    STACK GUARD            0000700000000000-0000700000001000 [    4K] ---/rwx SM=NUL  stack guard for thread 1

Thread 0:: Dispatch queue: com.apple.main-thread
0   libsystem_kernel.dylib          0x00007fff8edf9db6 __psynch_cvwait + 10
1   libsystem_pthread.dylib         0x00007fff8f04b728 _pthread_cond_wait + 767
2   libc++.1.dylib                  0x00007fff951af68f std::__1::condition_variable::wait(std::__1::unique_lock<std::__1::mutex>&) + 47
3   libmxnet.so                     0x0000000110e6a7d3 mxnet::engine::ThreadedEngine::WaitForVar(mxnet::engine::Var*) + 563
4   libmxnet.so                     0x0000000110ed86e7 mxnet::NDArray::SyncCopyToCPU(void*, unsigned long) const + 1111
5   libmxnet.so                     0x0000000110e3769d MXNDArraySyncCopyToCPU + 13
6   _ctypes.so                      0x00000001100bf7f7 ffi_call_unix64 + 79
7   _ctypes.so                      0x00000001100c0030 ffi_call + 840
8   _ctypes.so                      0x00000001100bb62a _ctypes_callproc + 591
9   _ctypes.so                      0x00000001100b5cad PyCFuncPtr_call + 1054
10  org.python.python               0x000000010f91c9b4 PyObject_Call + 99
11  org.python.python               0x000000010f99be30 PyEval_EvalFrameEx + 26668
12  org.python.python               0x000000010f99540e PyEval_EvalCodeEx + 1617
13  org.python.python               0x000000010f99fc9e fast_function + 117
14  org.python.python               0x000000010f99bf09 PyEval_EvalFrameEx + 26885
15  org.python.python               0x000000010f99fd31 fast_function + 264
16  org.python.python               0x000000010f99bf09 PyEval_EvalFrameEx + 26885
17  org.python.python               0x000000010f99540e PyEval_EvalCodeEx + 1617
18  org.python.python               0x000000010f994db7 PyEval_EvalCode + 48
19  org.python.python               0x000000010f9b883c run_mod + 53
20  org.python.python               0x000000010f9b88df PyRun_FileExFlags + 133
21  org.python.python               0x000000010f9b8430 PyRun_SimpleFileExFlags + 702
22  org.python.python               0x000000010f9c9d9e Py_Main + 3094
23  libdyld.dylib                   0x00007fff932965ad start + 1

Minimum reproducible example

No just run the example in the source code

Steps to reproduce

1. Install mxnet with setup in the official document: http://mxnet.io/get_started/osx_setup.html#install-mxnet-for-python

2. Download image_segementation pre trained model from baidu.yun

   FCN8S_VGG16_0019

3. Run example image_segmentation.py

   Segementation fault: 11

[piiswrong]

I get this error: [23:57:33] src/nnvm/legacy_json_util.cc:153: Loading symbol saved by previous version v0.8.0. Attempting to upgrade... [23:57:51] /Users/junyuanx/work/local/mxnet/dmlc-core/include/dmlc/./logging.h:300: [23:57:51] src/operator/./crop-inl.h:117: Check failed: param_.offset[0] <= data_shape[2]-out_shape[2] (34 vs. 33) offset[0] should be less than the residual space of height

seems to be the padding shape issue.
@tornadomeet could you update it?

[piiswrong]

@howard0su

[shadowinlife]

Thanks, Where Can I download compiled 0.8.0 mxnet, and have test on this pre-trained model?

[yajiedesign]

This issue is closed due to lack of activity in the last 90 days. Feel free to reopen if this is still an active issue. Thanks!

[huyangc]

Same error here.

---

I fixed it by change the crop offset of the last crop layer from 34 to 31 (according to the author's caffe github). I am now a little confused with the crop offset.

[szha]

@apache/mxnet-committers: This issue has been inactive for the past 90 days. It has no label and needs triage.

For general "how-to" questions, our user forum (and Chinese version) is a good place to get help.

[vandanavk]

Issue not reproducible with MXNet 1.3.1, Python 2.7 and Python 3.6.
Verified on DL AMI with Ubuntu.

@shadowinlife Can this issue be closed?

cc @sandeep-krishnamurthy

[zhreshold]

@vandanavk The original issue started with "Operating System: OS X 10.11.6". But you are testing on ubuntu.

[vandanavk]

Update: Tested on MacOS 10.12.6, MXNet 1.3.1, Python 2.7.15 too. Segmentation fault not observed.

Executed python2 image_segmentaion.py --input VOC2012/JPEGImages/2007_000027.jpg

@zhreshold thanks :) updated

[piyushghai]

@sandeep-krishnamurthy Requesting this issue to be closed as the Segmentation Fault is no longer observed.

@shadowinlife Please feel free to reopen if closed in error.
Thanks!