Open
Conversation
ilan-gold
reviewed
Feb 25, 2026
There was a problem hiding this comment.
Spent some time with this, seems like I would have some broad questions:
- Is it worth it to abstract the "math" from the "jax" via
array-api? My comments are along these lines. My software engineering instincts are "yes" but I don't know Jax or the math well enough to say either way and also it's a massive lift as it would require unifying with the current implementation. I'll try to spend some time reading up - In general, I see a few uses of things that have multiple implementations in
jax, specifically where one implementation lives injax.spicyand the other injax.numpy- It seems there are good reasons for this usually like slightly different features. But that raises the question further from the the above of how much of this could be refactored to be reused (modulo whichscipymodule is called for anumpyorjaxarray).
Not sure I'm the one most qualified to review, but figured I'd get the ball rolling! Maybe we should discuss among core what to do.
| return beta_, mu_, converged_ | ||
|
|
||
|
|
||
| def _compute_hat_matrix_diagonal( |
There was a problem hiding this comment.
| ) | ||
|
|
||
|
|
||
| def _compute_mu( |
There was a problem hiding this comment.
| w = mu / (1 + mu * alpha) | ||
| mat = (design_matrix.T * w) @ design_matrix | ||
|
|
||
| # There are a few ways to compute the log determinant. Here we use an approach |
There was a problem hiding this comment.
Why not https://docs.jax.dev/en/latest/_autosummary/jax.numpy.linalg.slogdet.html? It seems like it's in the array-api: https://data-apis.org/array-api/latest/extensions/generated/array_api.linalg.slogdet.html#slogdet and also curious if it has the same issue as the torch one?
| return n * dispersion_neg1 * jnp.log(dispersion) - logbinom.sum() | ||
|
|
||
|
|
||
| def _nb_loss_variable_terms( |
There was a problem hiding this comment.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Implement a
JaxInferencebackend for DESeq2 inference.This backend allows hardware acceleration via GPU/TPU. Please see the class docstring for a detailed description of the methodology.
A test file has been added to test all the same use cases as the default inference class and all tests pass.
Based on the import structure, the class can only be imported if jax is installed, so no added logic was implemented for handling missing imports.