I am executing somebody’s notebook in Jupyter. During a long calculation, I interrupt the cell, and the kernel restarts. All modules must be recompiled. How can I avoid this recompilation? Why should it be necessary if the precompiled files can be found under $HOME/.julia somewhere? Thanks.
I still have this issue, specifically when waiting for
Flux.jl computations in Jupyter. Interrupting (e.g., to terminate an indefinite training algorithm) will often kill the kernel, losing state, which is prohibitively annoying.
How can we fix kernel interruptions being fatal?