652994331 commented on issue #4464: URL: https://github.com/apache/incubator-tvm/issues/4464#issuecomment-673897424
@tobegit3hub Hi guys, i built tvm before and i built tvmsoop separately(not from USE_TF_TVMSOOP=ON) follow this:https://github.com/tobegit3hub/tftvm/tree/master/examples . After i got libxx.so and link them to my tvm home, i run the test import tensorflow as tf from tvm.contrib import tf_op mod = tf_op.Module("tvm_addone_dll.so") addone = mod.func("addone", output_shape=[4]) with tf.Session() as sess: a = tf.constant([10.1, 20.0, 11.2, -30.3]) b = addone(a) print(sess.run(b)) and i got this error: Traceback (most recent call last): File "test_python.py", line 5, in <module> addone = mod.func("add_one, output_shape=[4]") File "/opt/cephfs1/asr/users/qizhou.huang/.local/lib/python3.6/site-packages/tvm-0.7.dev1-py3.6-linux-x86_64.egg/tvm/contrib/tf_op/module.py", line 27, in func return Func(self.lib_path, name, output_dtype, output_shape) File "/opt/cephfs1/asr/users/qizhou.huang/.local/lib/python3.6/site-packages/tvm-0.7.dev1-py3.6-linux-x86_64.egg/tvm/contrib/tf_op/module.py", line 55, in __init__ self.module = load_library.load_op_library('tvm_dso_op.so') File "/opt/cephfs1/asr/users/qizhou.huang/anaconda3/envs/tvm/lib/python3.6/site-packages/tensorflow_core/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: /opt/cephfs1/asr/users/qizhou.huang/qizhou/PycharmProjects/incubator-tvm/build/tvm_dso_op.so: undefined symbol: _ZN10tensorflow12OpDefBuilder4AttrESs my gcc is 6.4.0,my tensorflow flow is tf-1.15.0 i use bazel build it from source and set -D_GLIBCXX_CXX11_ABI=1. Btw, i also tried the pip install tensorflow,-1.13.1, same error. Couild you please help me out, thanks ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected]
