exojax

exojax copied to clipboard

Release v1.2

14

Release branch for [v1.2](https://github.com/HajimeKawahara/exojax/milestone/3). I will merge this branch to `master` and release v1.2 in **November 21st** unless any serious issues are found. Comments and reviews welcome for anyone! -...

HajimeKawahara

Theorets I/O

2

I'd like to make the THEORETS I/O maybe in radis.api. https://theorets.tsu.ru/

HajimeKawahara

Vaexization of some algorithm

2

[Commented](https://github.com/HajimeKawahara/exojax/issues/314#issuecomment-1309887464) by @erwanp. Consider replacing some parts in PreMODIT to use Vaex directly. I need to get a little more familiar with Vaex though.

HajimeKawahara

related to https://github.com/HajimeKawahara/exojax/pull/313#issuecomment-1321363332 and https://github.com/HajimeKawahara/exojax/pull/313#discussion_r1027280614, we need to consider how we control device memory use when calling api.Mdb.

HajimeKawahara

Premodit Elower grid error rethinking

6

Previously, I estimated a maximum error of delta E_lower. But, it looks too pessimistic. To reduce more memory use, we need to rethink it.

HajimeKawahara

PretiMODIT: inverse T-grid instead of Elower

1

I feel that the inverse T grid should be tried as one of the LBD grids instead of Elower. tag: Precomputation+T inverse+MODIT ### maybe good - less sparse than E-grid...

HajimeKawahara

wavenumber dependent IP

1

For a wide range data, we need a wavenumber dependent IP convolution to spectra.

HajimeKawahara

radiative transfer rethinking

2

In Paper I, we used 2 E_3 as a pure absorption transmission T, but not tried a numerical integral of mu e^-dtau/mu, for instance, using Gaussian quadrature etc, which might...

HajimeKawahara

During implementation of PreMODIT, I found that jax.lax.scan in modit.xsmatrix can be replaced by vmap. But not sure everything is OK and/or vmap is better than scan. I will check...

HajimeKawahara

GPU dependence and non-reproducibility of radiative transfer (small error)

1

During the review of #265 , we found that there is non-reproducibility of radiative transfer results for MODIT and PreMODIT. This can be confirmed by the unit tests: - exojax/tests/unittests/spec/modit/modit_test.py...

HajimeKawahara