Thibault Monsel

Results 23 comments of Thibault Monsel

Thanks for the review Patrick :) , so from what I understood what i need to do is to create a new class `DDEImplicitNonLinearSolve` (for e.g) that inherits from `AbstractNonlinearSolver`...

Gotcha, I might be running into an issue with regards to `implicit_jvp`. I think I need to rewrite the `_rewrite(root, _, diff_args, closure)` function since it returns `fn(root,args)` and `fn`...

Ok so this is what I came up with by transforming the fixed point problem in a root finding one. ``` py def history_extrapolation_implicit(terms, delays, dense_info, dense_interp, solver, direction, t0,...

With the current setup > ```python > def history_extrapolation_implicit(terms, delays, dense_info, dense_interp, > solver, direction, t0, tprev, tnext, y0_history): > > nonlinearsolver = FixedPointSolver() > nonlinear_args = (terms,implicit_step,dense_interp,solver,delays,t0,y0_history,state,args) > >...

I added some auxiliary stats for implicit and explicit counting (https://github.com/patrick-kidger/diffrax/pull/169#pullrequestreview-1194907612). I also did some documentation and testing. Documentation will most likely change later on. Let me know if I...

Hi @patrick-kidger , I'm having another go on the DDEs and with all of the changes on Diffrax and Equinox. I have rebased the code and with it, there is...

Adding `[...]` did the trick for `ts` ```py ts = state.dense_ts[...] ``` Unfortunately unwrapping the Buffer with other structure's like `DenseInfos` didn't seem to work. ```py infos=jtu.tree_map(lambda x: x[...], state.dense_infos)...

Indeed your right, I just realised that in the documentation yesterday, this makes sense now ! However, we only have access to `dense_info` after creating `DenseInterpolation`. i.e. ```py dense_interp =...

Sorry about that, lmk if this is clearer now. For the type it says : ``` TypeError: cache_clear(): incompatible function arguments. The following argument types are supported: E 1. (self:...

This seems to reappear in `pip==23.0.1` with the opened issue : https://github.com/google/jax/issues/15387. Solution : By adding `--no-build-isolation` in the CLI did the trick for me. eg `python3 -m pip install...