Replies: 3 comments
-
Hi - can you say more about what your "typical jax script" includes? This warning (and the deadlock that it warns about) will occur if your script attempts to use JAX together with multiprocessing. JAX's computational model is not compatible with multiprocessing. Is it possible your script is attempting to use JAX with some library like |
Beta Was this translation helpful? Give feedback.
-
One example of how this comes up is if you run a command in a jupyter notebook, using !, e.g. !ls -alh veryintriguingpath Or if you use imageio to write an mpeg file. Many packages seem to fork as a matter of course when calling subprocess. It would be nice if Jax didn't rebel when you need to invoke such packages. But tbh how it all works is honestly quite beyond me, so I'm not sure how/if this can be done. |
Beta Was this translation helpful? Give feedback.
-
Well, it's a warning. You are free to ignore it and things might work if you get lucky. It is unfortunately correct in this case: calling The warning issued on a jupyter The right way to fix that problem would be upstream in Hope that helps! |
Beta Was this translation helpful? Give feedback.
-
I run a typical
jax
script and this warning pops upafter running
import jax; jax.print_environment_info()
I getThis was after a fresh reinstall in a new virtual environment. Has anyone got any pointers here? I assume this works for other people and I'm doing something wrong!
Beta Was this translation helpful? Give feedback.
All reactions