diffrax

diffrax copied to clipboard

For each adjoint method, it would be convenient to include a table specifying which are compatible with each kind of autodifferentiation.

patrick-kidger

Can't return solution of coupled differential equations

5

I'm trying to solve a mid-sized system of coupled differential equations with `diffrax`. I'm using version 0.2.0. Here's a short snippet of dummy code that raises the issue I'm having:...

cgiovanetti

Adding support for complex dtypes

2

Hi, I finally have the time to actually finish this PR. I've been trying to start from scratch instead of going over all of the changes, and currently the solvers...

golanor

Memory growing when integrating lots (different) functions

3

Hi again 😃 I noticed that memory seems to grow indefinitely when integrating lots of newly defined functions. In the following example, memory grows at each iteration if `vector_field` is...

fhchl

At the moment it uses `jnp.linalg.solve`, which assumes that the solve operation is well-posed. This needn't actually be the case, though, and at the moment the gradient we obtain could...

patrick-kidger

Error When I use TPU on Google Colab.

3

I try to use TPU on Colab with Diffrax. Error appears when I use TPU but no problem with CPU or GPU. ``` import jax.tools.colab_tpu from diffrax import diffeqsolve, Dopri5,...

tentechit

Add differentiable event handling

4

One thing that needs thinking about is how this compares to the use of jump processes as a driving control, which morally speaking do something very similar. - [x] Discrete...

patrick-kidger

At the moment, padding the state with constant values (zero vector field) means that `rms_norm(state) -> 0` as this padding increases. This can cause issues with steady state detection, and...

patrick-kidger

Solving for initial value conditions and final value conditions

10

Hey @patrick-kidger , I recently started exploring diffrax (just realised how powerful diffrax is!, and kudos for the amazing work!). I was thinking if there was a way to solve...

enigmarikki

In this case the size of the output array should be equal to the known-at-compile-time number of steps.

patrick-kidger