tensorflow.python.framework.errors_impl.NotFoundError

[chuchienshu]

python3 -m src.flownet2.test --input_a data/samples/0img0.ppm --input_b data/samples/0img1.ppm --out ./
Traceback (most recent call last):
  File "/usr/lib/python3.5/runpy.py", line 184, in _run_module_as_main
    "__main__", mod_spec)
  File "/usr/lib/python3.5/runpy.py", line 85, in _run_code
    exec(code, run_globals)
  File "/home/chuchienshu/tensorflowcode/newflownet/flownet2-tf/src/flownet2/test.py", line 4, in <module>
    from .flownet2 import FlowNet2
  File "/home/chuchienshu/tensorflowcode/newflownet/flownet2-tf/src/flownet2/flownet2.py", line 2, in <module>
    from ..flownet_css.flownet_css import FlowNetCSS
  File "/home/chuchienshu/tensorflowcode/newflownet/flownet2-tf/src/flownet_css/flownet_css.py", line 2, in <module>
    from ..flownet_cs.flownet_cs import FlowNetCS
  File "/home/chuchienshu/tensorflowcode/newflownet/flownet2-tf/src/flownet_cs/flownet_cs.py", line 2, in <module>
    from ..flownet_c.flownet_c import FlowNetC
  File "/home/chuchienshu/tensorflowcode/newflownet/flownet2-tf/src/flownet_c/flownet_c.py", line 3, in <module>
    from ..correlation import correlation
  File "/home/chuchienshu/tensorflowcode/newflownet/flownet2-tf/src/correlation.py", line 4, in <module>
    tf.resource_loader.get_path_to_datafile("./ops/build/correlation.so"))
  File "/home/chuchienshu/.local/lib/python3.5/site-packages/tensorflow/python/framework/load_library.py", line 64, in load_op_library
    None, None, error_msg, error_code)
tensorflow.python.framework.errors_impl.NotFoundError: /home/chuchienshu/tensorflowcode/newflownet/flownet2-tf/src/./ops/build/correlation.so: undefined symbol: _ZN10tensorflow15shape_inference16InferenceContext8WithRankENS0_11ShapeHandleExPS2_

[jakeoung]

Same problem.

[sampepose]

What OS, gcc version, nvcc version? Did the make process have any warnings or errors? What GPU do you have? Also, I’ve never tried with Python 3, try using Python 2.

I’ll try to repro on Monday once I get that info.

On Sep 12, 2017, at 7:40 PM, chuchienshu notifications@github.com wrote:

python3 -m src.flownet2.train Traceback (most recent call last): File "/usr/lib/python3.5/runpy.py", line 184, in _run_module_as_main "main", mod_spec) File "/usr/lib/python3.5/runpy.py", line 85, in _run_code exec(code, run_globals) File "/home/chuchienshu/tensorflowcode/newflownet/flownet2-tf/src/flownet2/train.py", line 1, in from ..dataloader import load_batch File "/home/chuchienshu/tensorflowcode/newflownet/flownet2-tf/src/dataloader.py", line 7, in tf.resource_loader.get_path_to_datafile("./ops/build/preprocessing.so")) File "/home/chuchienshu/.local/lib/python3.5/site-packages/tensorflow/python/framework/load_library.py", line 64, in load_op_library None, None, error_msg, error_code) tensorflow.python.framework.errors_impl.NotFoundError: /home/chuchienshu/tensorflowcode/newflownet/flownet2-tf/src/./ops/build/preprocessing.so: cannot open shared object file: No such file or directory — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.

[jakeoung]

It seems that tensorflow-gpu 1.3.0 version doesn't work to me. I managed to run the code. For your information, my platform that can run is the following:

tensorflow-gpu (1.0.1) CUDA 8.0 Ubuntu 14.04

[HuachunWang]

You can try it under configurations: gcc4.9, python2.7, Tensorflow1.2.0, cuda8.0, cudnn5.1, GTX TitanX. I have tried Tensorflow1.1.0 and Tensorlfow1.0.1, both did't work. Only Tensorflow1.2.0 did work!

[msokoloff1]

Comment out this part of line 56 in the makefile : "-D_GLIBCXX_USE_CXX11_ABI=0". That works for Tensorflow 1.3, cuda 9, titan v

[xinyangrobotic]

the same questions with you

[chenguanqi]

You can enter "make" command in the terminal under the path of Makefile. Then it will generate the files in need. ~/flownet2-tf$ make

[yinjunbo]

@zxy-HIT @ChenGQ26
Hi！Could you please tell me have you succeed in training?

[chenguanqi]

@yinjunbo Hi! I didn‘t try it in training.

[liruiw]

Hi! I have the same issue and didn't get resolved by commenting out that line. I'm using tf1.7, cuda9.0 python2.7, gcc5.4 and TitanX as GPU. Thanks!

[ceciliavision]

Make sure you used the same python version to compile (specified in Makefile) and run.

[clausmichele]

same problem with Python 2.7, Tensorflow 1.4, Cuda 8.0, Ubuntu 16.04. Tried to comment that line before building but nothing changed.

[mithun2595]

I faced the same issue with python3, tf 1.5. Fixed it by following this issue https://github.com/tensorflow/tensorflow/issues/13607

[gongbudaizhe]

TensorFlow 1.2.1 also works, but not > 1.2.1

[aa-samad]

facing the same issues, finally got results #28