It seems that TensorFlow binaries probably mismatch platform.
I have this output:
Sortie d’erreur :
2017-11-29 12:21:41.546619: I tensorflow/core/platform/cpu_feature_guard.cc:137] Your CPU supports instructions that this TensorFlow binary was not compiled to use: SSE4.1 SSE4.2 AVX
MemoryError
at input_data.py. in next_batch on line 146 at Answer.py. in <module> on line 39
Please ignore the first warning message (this warning message is new and does not impact the rest of the code). Please check with the new default code for this puzzle (save your code, reset code, compare), you will see the lines that remove this warning message.
As for the MemoryError, the new TensorFlow version consumes much more memory than it used too. Please try to lower the size of your data.
If you cannot make it work due to this memory issue (or you can make it work but then you cannot pass the test), let us know.
The new default code says uncomment to get a prediction. But when I uncomment, all I get is a compilation error with sess and y undefined. Why do u put such code if it won’t compile?