[sylvesterkaczmarek]

Problem

rgb_to_hsl masks achromatic hue/saturation outputs after dividing by the zero colour range. The forward value for a grey pixel is correct, but reverse differentiation produces NaNs, including in the lightness component. Low-precision black/white pixels also need a safe denominator for the inactive saturation branch.

Change

Replace inactive denominators before division, preserving the existing achromatic output branches and chromatic formula. This follows the JAX guidance on masked undefined operations.

Validation

• Earlier full colour-conversion run: 171 passed, 40 skipped (JAX 0.7.0, TensorFlow 2.20.0).
• Current rebased head: 9 HSL regressions and 20 adjacent conversion tests passed; 5 skipped (JAX 0.12.0, TensorFlow 2.21.0). Tests now live in dm_pix/_src/color_conversion_test.py. Unchanged production code fails eight cases and passes the chromatic finite-difference control.
• Coverage includes black/grey/white, float16/bfloat16/float32, eager/JIT execution, forward/reverse differentiation along the greyscale path, mixed images, and chromatic gradients compared with independent colorsys finite differences.
• Python syntax and git diff --check pass on the current head.

Hue is mathematically undefined on the achromatic axis. The tests preserve the existing constant hue/saturation branches and verify the lightness derivative along the greyscale path; they do not assert a unique full colour-space derivative there.

Python 3.12.14 on macOS arm64 CPU. GPU/TPU and the complete image-processing suite were not run. Channel-axis handling in #125 is separate and unchanged. No epsilon, colour convention, dependency, or public API changes.

Reproduced against ffb65074b39b67e1a686d1348f2d461c7ff3ab45.

[claudiofantacci]

Please move these tests under color_conversion_test.py. We already have a color transformation test, so you can put them there, thanks!

[sylvesterkaczmarek]

Moved the tests into color_conversion_test.py. All 9 new cases and 20 adjacent tests pass on current master.

[claudiofantacci]

Thanks!

[pabloduque0]

Left a couple nits, otherwise LGTM

[sylvesterkaczmarek]

Thanks for the review, Pablo. I addressed both nits in 73f8b59: the float64 setup is now in one line, and the mixed-image test separates arrange, act and assert. The 9 focused HSL-gradient tests still pass on CPU.

[claudiofantacci]

@sylvesterkaczmarek something went wrong with merging back some changes I did internally with yours. I guess the code was at an old commit, sorry about that. Can you please look at the current code again? I'll then redo a review, see internally what happens and hopefully we'll close the loop on this PR.

[sylvesterkaczmarek]

Thanks for flagging this, Claudio. I checked current master against the PR branch: the HSL fix and regressions are already on master via 99ad28a. My sync inadvertently removed x from the finite-difference test, which explains the Python CI NameError. I've restored it in 2881e23 (syntax and diff checks pass). Happy to close this as already integrated if that suits your internal workflow. Appreciate your help getting this across the line.