I tried to lightly use the machine learning library TensorFlow http://tensorflow.org/ released by Google on 11/9, and organized the warnings and errors that I encountered, which seem to have not been released in Japanese yet. I will do it.
As for how to use it, there is a polite tutorial in the head family, and other articles have tried to translate it into Japanese, so please refer to that.
The author's environment is OSX 10.9.5 and uses binaries for MacOS.
>>> import tensorflow as tf
module compiled against API version 9 but this version of numpy is 6
It seems that the old numpy that was already in is referenced
If you rename numpy in numpy. \ _ \ _ Path \ _ \ _ appropriately such as numpy_tmp, import tensorflow will pass. (If you have better know-how as to whether it is okay to deal with it with rename, please teach the python shop.)
>>> import numpy
>>> numpy.__version__
You should get an old version of numpy, because the above error says numpy is 6'1.6.0'Such
>>> numpy.__path__
You should see the path of numpy you are looking at now, write it down and rename it
>>> sess = tf.Session()
can't determine number of CPU cores: assuming 4
I tensorflow/core/common_runtime/local_device.cc:25] Local device intra op parallelism threads: 4
http://stackoverflow.com/questions/33617638/tensorflow-mac-os-x-cant-determine-number-of-cpu-cores According to this, this seems to be happening when using OSX binaries,
NUM_CORES = <Number of cores>
If you explicitly specify, it will not appear. It is supported by the commit of https://github.com/tensorflow/tensorflow/commit/430a054d6134f00e5188906bc4080fb7c5035ad5, and it seems that it will be executed without problems in the binary released in the future.
If you encounter anything else, I will add it. (I'm not usually a python man, so if there is something I don't understand, I may not understand it well. I would appreciate it if you could point out and teach me.)
Recommended Posts