terminated,exit value:139 (Segmentation fault (core dumped)) #5
Comments
|
I my opinion, it could because of 2 reasons:
|
|
@mjiUST Thanks
Sadly, there is no error log, just the python interpreter stoped.(<terminated, exit value: 139>main.py [debug] [/home/cd/miniconda2/envs/SurfaceNet/bin/python2.7]) patches_embedding[:,:] = patch2embedding_fn(patch_allBlack)[0]# don't use np.repeat (out of memory) It may cause by a array problem, patch2embedding_fn is a suspicions array. |
|
kindly note that the output of command
From your printout "...Using cuDNN version 5005 on context None...", we infer that the cudnn version (5.1) is not correctly linked to your theano. Please make sure the cudnn in the If your machine have the following configuration:
I recommend you locate and delete the zombie cudnn (5.0) before continue. You can checkout the output of the commands |
|
@mjiUST Thanks [Thread debugging using libthread_db enabled] rtld_fini=, stack_end=0x7ffef436cbd8) at ../csu/libc-start.c:291 #135 0x000055aed3a7b87f in _start () |
|
@cdb0y511
Let's see what we can do. |
|
@mjiUST Thanks Your programme has just finished(middbury Dino). It takes about 38mins(fast_run). The nets is loaded from 2D_2_3D-19-0.918_0.951.model. |
|
@cdb0y511
Yes, your running time is similar to mine.
Yes, the training and validation split are from the DTU dataset. And it generalize well on the Dino model. |
After I follow this #4, the enviroment problem seems to be gone.
But evertime python main.py occour Segmentation fault (core dumped), the python interpreter stops on main_reconstruc.py line 73
patches_embedding, inScope_cubes_vs_views = earlyRejection.patch2embedding(
images_list, img_h_cubesCorner, img_w_cubesCorner, patch2embedding_fn, patches_mean_bgr,
N_cubes, N_views, D_embedding, patchSize = params.__imgPatch_hw_size,
batchSize = params.__batchSize_similNet_patch2embedding,
cubeCenter_hw = np.stack([img_h_cubesCenter, img_w_cubesCenter], axis=0)) # (N_cubes, N_views, D_embedding), (N_cubes, N_views).
output like this
start reconstruction ...
loaded img ./inputs/Middlebury/dinoSparseRing/dinoSR0007.png
loaded img ./inputs/Middlebury/dinoSparseRing/dinoSR0008.png
loaded img ./inputs/Middlebury/dinoSparseRing/dinoSR0009.png
loaded img ./inputs/Middlebury/dinoSparseRing/dinoSR0010.png
loaded img ./inputs/Middlebury/dinoSparseRing/dinoSR0011.png
loaded img ./inputs/Middlebury/dinoSparseRing/dinoSR0012.png
xyz bounding box of the reconstructed scene: [-0.061897 0.010897], [-0.018874 0.068227], [-0.057845 0.015495]
loaded similarityNet model: ./inputs/SurfaceNet_models/epoch33_acc_tr0.707_val0.791.model
loaded SurfaceNet model: ./inputs/SurfaceNet_models/2D_2_3D-19-0.918_0.951.model
Segmentation fault (core dumped)
It seems like a stack or memery problem. Can you help me? Any advice?
My OS is
Ubuntu 1604 x64
cuda 8.0+ cudnn5.1 and cuda 9.0+cudnn7.1.1
1080Ti
.theanorc
[global]
floatX=float32
device=cuda
optimizer=None
allow_gc=True
gpuarray.preallocate=0.75
gcc.cxxflags=-Wno-narrowing
exception_verbosity=high
[lib]
cnmem=0.75
[nvcc]
nvcc.fastmath=True
[cuda]
root=/usr/local/cuda-8.0
@mjiUST @Rubikplayer
The text was updated successfully, but these errors were encountered: