search

XuyangBai / D3Feat

[TensorFlow] Official implementation of CVPR'20 oral paper - D3Feat: Joint Learning of Dense Detection and Description of 3D Local Features https://arxiv.org/abs/2003.03164
MIT License
261 stars 38 forks source link

Demo problem #9

Closed junzastar closed 3 years ago

junzastar commented 4 years ago

Sorry to bother. I have a question about demo_registration.py. When I run it, it always has the "Segmentation fault" problem. What is wrong with my device? The code was running on cuda10.0, Tensorflow1.12, Tesla M40. The log is as follows: Instructions for updating: keep_dims is deprecated, use keepdims instead 2020-06-21 13:49:59.269588: I tensorflow/core/platform/cpu_feature_guard.cc:141] Your CPU supports instructions that this TensorFlow binary was not compiled to use: SSE4.1 SSE4.2 AVX AVX2 FMA 2020-06-21 13:50:03.450215: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1432] Found device 0 with properties: name: Tesla M40 major: 5 minor: 2 memoryClockRate(GHz): 1.112 pciBusID: 0000:02:00.0 totalMemory: 11.18GiB freeMemory: 11.07GiB 2020-06-21 13:50:03.513556: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1432] Found device 1 with properties: name: Tesla M40 major: 5 minor: 2 memoryClockRate(GHz): 1.112 pciBusID: 0000:82:00.0 totalMemory: 11.18GiB freeMemory: 11.07GiB 2020-06-21 13:50:03.513649: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1511] Adding visible gpu devices: 0, 1 2020-06-21 13:50:04.343520: I tensorflow/core/common_runtime/gpu/gpu_device.cc:982] Device interconnect StreamExecutor with strength 1 edge matrix: 2020-06-21 13:50:04.343572: I tensorflow/core/common_runtime/gpu/gpu_device.cc:988] 0 1 2020-06-21 13:50:04.343579: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1001] 0: N N 2020-06-21 13:50:04.343584: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1001] 1: N N 2020-06-21 13:50:04.344708: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1115] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 10709 MB memory) -> physical GPU (device: 0, name: Tesla M40, pci bus id: 0000:02:00.0, compute capability: 5.2) 2020-06-21 13:50:04.345655: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1115] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:1 with 10709 MB memory) -> physical GPU (device: 1, name: Tesla M40, pci bus id: 0000:82:00.0, compute capability: 5.2) End of train dataset

self.neighborhood: [37 30 34 36 37] 2020-06-21 13:50:08.567410: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1511] Adding visible gpu devices: 0, 1 2020-06-21 13:50:08.567557: I tensorflow/core/common_runtime/gpu/gpu_device.cc:982] Device interconnect StreamExecutor with strength 1 edge matrix: 2020-06-21 13:50:08.567571: I tensorflow/core/common_runtime/gpu/gpu_device.cc:988] 0 1 2020-06-21 13:50:08.567578: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1001] 0: N N 2020-06-21 13:50:08.567584: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1001] 1: N N 2020-06-21 13:50:08.568705: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1115] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 10709 MB memory) -> physical GPU (device: 0, name: Tesla M40, pci bus id: 0000:02:00.0, compute capability: 5.2) 2020-06-21 13:50:08.568985: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1115] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:1 with 10709 MB memory) -> physical GPU (device: 1, name: Tesla M40, pci bus id: 0000:82:00.0, compute capability: 5.2) Model restored from results/Log_contraloss/snapshots/snap-54 Segmentation fault

Can you give me some advice? Thx.

XuyangBai commented 4 years ago

Hi @ZhoujunSKY Sorry I haven't tried our method on Tesla M40, so it's really hard to say what's the reason for the segmentation fault. Could you check if the program will crash for other TensorFlow models first? There are quite a lot of reasons for TensorFlow to raise segmentation fault on their Github issues.

junzastar commented 4 years ago

Thanks for your response. Actually, I am following your instructions by using conda to create a virtual environment (environment.yml). Is the version of TensorFlow wrong?

Hi @ZhoujunSKY Sorry I haven't tried our method on Tesla M40, so it's really hard to say what's the reason for the segmentation fault. Could you check if the program will crash for other TensorFlow models first? There are quite a lot of reasons for TensorFlow to raise segmentation fault on their Github issues.

XuyangBai commented 4 years ago

I use tf 1.12 and cuda 9.0 for 2080Ti. I haven't met the segmentation fault under my environment so I have no much idea about this error. Could you first verify whether this tf version is compatible with your cuda and gpu?

junzastar commented 4 years ago

Thanks a lot. I will debug my TF environment.

tracykim commented 4 years ago

Hi @ZhoujunSKY , I had same problem like this and solve the problem in the following ways:

import tensorflow first and then import open3d. or you can see this issue

good luck to you!