Rémi Louf

This looks like the dispatcher wasn't updated after a scan refactor. I'll give it a shot using the current Numba dispatcher as a reference.

Can we close this as _Not Planned_?

There shouldn't be anything to implement here, the idea is to add an alias to get Aesara's API closer to numpy's. Or better, replace the use of `mul` by `multiply`,...

If I understand well, for now `Aeasara` provides `RandomStateType` and `RandomGeneratorType` that are wrappers around `numpy.random.RandomState` and `numpy.random.Generator` respectively. However the API doesn't follow numpy's as we have to pas...

Thank you for opening an issue. Do you still observe this behavior after setting the environment variable `PYTHONHASHSEED` to e.g. `0`? See [the python documentation](https://docs.python.org/3/using/cmdline.html#envvar-PYTHONHASHSEED).

> Hello, `PYTHONHASHSEED` does indeed fix this, although it's not ideal since we want to be able to keep the benefits of randomized hashing (eg. avoid dictionary keying attacks). Thank...

> In addition, would Aesara still be able to generate JAX code that we can pass around to other libraries (e.g., BlackJax which will call grad/JIT on a user defined...

To follow up on the previous discussion. We were considering the following function: ```python from jax import lax from jax import make_jaxpr import jax.numpy as jnp def add_fn(a, b): return...

~~Yes.~~ `jax.grad` requires tracing the function to build a "new graph" so it will not be possible to pass as an argument a function that is built this way. It...

It is clear now that by targeting JAX's IR directly we would still be able to use `jax.jit`, `jax.vmap` on the compiled function, but will not be able to apply...