Error dp_train

[vijendra037]

I have installed deepmd kit according to instruction but with tensorflow version 1.13 and python 3.7.3.
while running dp_train file I got error. I checked your troubleshoot and again install with -DTF_GOOGLE_BIN=true in the process of cmake.
but again I get this error.

Traceback (most recent call last):
File "/home/vkumar/tensorflow_env_dir/deepmd_root/bin/dp_train", line 16, in
from deepmd.Model import NNPModel
File "/home/vkumar/tensorflow_env_dir/deepmd_root/bin/../lib/deepmd/Model.py", line 21, in
op_module = tf.load_op_library(module_path + "libop_abi.so")
File "/home/vkumar/anaconda3/envs/tensorflow_env/lib/python3.7/site-packages/tensorflow/python/framework/load_library.py", line 61, in load_op_library
lib_handle = py_tf.TF_LoadLibrary(library_filename)
tensorflow.python.framework.errors_impl.NotFoundError: /home/vkumar/tensorflow_env_dir/deepmd_root/lib/deepmd/libop_abi.so: undefined symbol: _ZN10tensorflow12OpDefBuilder4AttrESs

please help me for this.

[njzjz]

Maybe your version of TensorFlow for Python is not consistent with that of TensorFlow for C++. Is your version of TensorFlow for C++ v1.13?

[dingye18]

It looks like the libop_abi.so did not find the right path to libtensorflow_framework.so. Did you add the $tensorflow_root/lib to your LD_LIBRARY_PATH environment parameters?
If not, try this command:
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$tensorflow_root/lib.
BTW, it is not suggested to use python3.7.3 for tensorflow. There will be some unexpected errors when use python 3.7.3 for tensorflow. Is python 3.7.x supported with Tensorflow

[Paul-St-Young]

I encountered a similar error, but via a different route, using Python 3.6.8 and Watson Machine Learning conda environment v1.6.1.

deepmd/__init__.py:2 -> from .DeepEval   import DeepEval
DeepEval.py:18 -> op_module = tf.load_op_library(os.path.join(module_path, "libop_abi.{}".format(ext)))
File "/opt/apps/anaconda3/envs/wmlce-v1.6.1-py3.6/lib/python3.6/site-packages/tensorflow/python/framework/load_library.py", line 61, in load_op_library
lib_handle = py_tf.TF_LoadLibrary(library_filename)
tensorflow.python.framework.errors_impl.NotFoundError:
.local/lib/python3.6/site-packages/deepmd/libop_abi.so: undefined symbol: _ZN10tensorflow12OpDefBuilder4AttrESs

[Paul-St-Young]

@dingye18 libtensorflow_framework.so is in my LD_LIBRARY_PATH.
nm libtensorflow_framework.so | grep _ZN10tensorflow12OpDefBuilder4 shows
0000000000680a98 t 0000007c.plt_call._ZN10tensorflow12OpDefBuilder4AttrENSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEE' '0000000000e94670 T _ZN10tensorflow12OpDefBuilder4AttrENSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEE

Basically, OpDefBuilder4AttrENSt7 is in my tensorflow library, but libop_abi.so is looking for OpDefBuilder4AttrESs.
I'm trying to use DeePMD 1.1.4 with tensorflow 1.14.0 distributed inside wmlce-v1.6.1.
Is there an expected incompatibility? How can I circumvent this error?

[njzjz]

@dingye18 libtensorflow_framework.so is in my LD_LIBRARY_PATH.
nm libtensorflow_framework.so | grep _ZN10tensorflow12OpDefBuilder4 shows
0000000000680a98 t 0000007c.plt_call._ZN10tensorflow12OpDefBuilder4AttrENSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEE' '0000000000e94670 T _ZN10tensorflow12OpDefBuilder4AttrENSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEE

Basically, OpDefBuilder4AttrENSt7 is in my tensorflow library, but libop_abi.so is looking for OpDefBuilder4AttrESs.
I'm trying to use DeePMD 1.1.4 with tensorflow 1.14.0 distributed inside wmlce-v1.6.1.
Is there an expected incompatibility? How can I circumvent this error?

Please check if you use a different gcc version for tensorflow and deepmd-kit (for example 4.8 and 5.4). Please use the same compiler version.