Dan Foreman-Mackey

@ColCarroll: I find that the truth of your statement is somewhat stochastic :D, but yes, sometimes restarting and running again works (but not always)! Edited to add: In my experience...

@Joshuaalbert — JAX doesn't currently offer a public API for this (customizing multiple transforms for a single callable), but it's on our radar. I'd say that this is off topic...

For the first one: make sure that you're using the most recent release of pip. For the second: does `rm -rf build; python setup.py develop` help? Do you get any...

Perhaps you already know this, but I normally use a hack to set the `compiledir` using `os.pid`. It's possible that this could also be handled by pre-compiling the required theano...

Something like the following can work: ```python import os from multiprocessing import Pool os.environ["THEANO_FLAGS"] = f"compiledir={os.getpid()}" import theano import theano.tensor as tt def func(x): x_ = tt.dscalar() return theano.function([x_], [x_...

Or... ```python from multiprocessing import Pool import theano import theano.tensor as tt if __name__ == "__main__": x_ = tt.dscalar() func = theano.function([x_], [x_ * x_]) with Pool(4) as pool: print(list(pool.map(func,...

Thanks @tyler-a-cox! It's great to know that you have a use case. One question, how important is GPU support? AFAIK, there still isn't any GPU support for Type 3 transforms...

Thanks @eelregit! I must admit that none of my use cases would really be counted as "Machine Learning". In fact, I don't have immediate applications, but I'd be excited to...

Great question! This is an issue introduced by this PR: https://github.com/exoplanet-dev/celerite2/pull/68 I should probably spend some time thinking about how to fix this properly, but one option for the short...

Yeah - that would be useful and the interface is very different from `celerite`. I'm not sure when I'll be able to get to this, but some combination of the...