kernel keeps dying in jupyter notebook

Determining why jupyter notebook kernel dies can be daunting sometimes. When kernel dies as a result of library issues, you might not get any feedback as to what is causing it. Try running the code that is causing the kernel to die in a terminal or in ipython. One common issue I have identified is the use of SSE4.1 in later versions of tensorflow, which gives this error: TensorFlow library was compiled to use SSE4.1 instructions, but these aren't available on your machine. Aborted (core dumped). This will cause the kernel to die if you are using tensorflow in your code and not display any error in notebook. If this is the case, uninstall tensorflow and install with conda: conda install tensorflow==1.12.0, for instance.

Leave a Comment