I'm using anaconda and every thing is a mass..
menglin0320 opened this issue · 5 comments
I'm trying to follow the instruction here
But I don't know where should I put the build_user_op.sh and call it if I'm using anaconda,
I found /home/menglin/anaconda2/envs/tensorflow_env/lib/python2.7/site-packages/tensorflow/include
is a pretty good choice
and I put your user_ops on
/home/menglin/anaconda2/envs/tensorflow_env/lib/python2.7/site-packages/tensorflow/include/ tensorflow/core/
But that doesn't work. And I don't know what to do now
I've not used anaconda, but I think it should be the same.
Clone this repository into any folder on your computer, and don't move anything. Simply run the build_user_op.sh file, and it should compile the ops.
Then, you just have to import these compiled files from your code. Take a look at my example here, in the second code block - see how I'm importing the .so files? Do the same thing, but provide the path to wherever on your system the compiled files are located.
NotFoundError Traceback (most recent call last)
in ()
5 # Import the forward op
6 roi_pooling_module = tf.load_op_library(
----> 7 home + "/anaconda2/envs/tensorflow_env/lib/python2.7/site-packages/tensorflow/include/tensorflow/core/user_ops/roi_pooling_op.so")
8 roi_pooling_op = roi_pooling_module.roi_pooling
9
/home/menglin/anaconda2/envs/tensorflow/lib/python2.7/site-packages/tensorflow/python/framework/load_library.pyc in load_op_library(library_filename)
69 return _OP_LIBRARY_MAP[library_filename]
70 # pylint: disable=protected-access
---> 71 raise errors._make_specific_exception(None, None, error_msg, error_code)
72 # pylint: enable=protected-access
73 finally:
NotFoundError: libcudart.so.8.0: cannot open shared object file: No such file or directory
I got this error
The problem is on the step running these two cammands
bazel build -c opt --config=cuda //tensorflow/core/user_ops:roi_pooling_op.so
bazel build -c opt --config=cuda //tensorflow/core/user_ops:roi_pooling_op_grad.so
in build_user_op.sh, I feel I should do it on the root of the tensorflow I'm using instead of do it on the Cloned repository. But I don't know where is the root since the lay out of tensorflow in anaconda is pretty different from normal tensorflow.
As I said I found the include folder highly possible to correspond to the root file on the cloned repository, but that doesn't work. I copied the script to the include file and tried to run it. But I got
The 'build' command is only supported from within a workspace.
The 'build' command is only supported from within a workspace.
So I'm very confused now
and my tensorflow version is 0.9 and I found the tensorflow you are using is likely to be 0.8, will that cause any trouble???
It works now, thanks for your help.