commit 92185c7be9e3a71b0112102af6f86e7d559bb65b
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Mon Jul 27 09:05:52 2026 +0200

dev-python/stop-sequencer: declare the pytorch dependency, revbump

Both modules in this package start with a bare `import torch`, so it
cannot be imported at all without pytorch. Upstream's setup.py declares
only transformers, and the ebuild inherited that omission.

Depending on transformers does not cover it. sci-ml/transformers has
IUSE="torch" with the flag off by default, so a default build of
transformers pulls no pytorch, and stop-sequencer would then fail on
import while its declared dependency was satisfied.

Depend on sci-ml/pytorch directly rather than forcing
transformers[torch]: the import here is of torch itself, and the USE
flag would additionally drag in accelerate and caffe2 that nothing in
this package uses. Both are single-impl, so the existing
$ form applies unchanged.

commit 0a7851b2265164a2a6f59ce454af15d8000df611
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sun May 10 18:38:02 2026 +0200

dev-python/stop-sequencer: switch to DISTUTILS_SINGLE_IMPL

sci-ml/transformers is SINGLE_IMPL; multi-impl consumer with bare
$ produces python_targets_python3_*(-)? that the
single-impl child can't expose, blocking emerge resolution under
multi-target profiles. Same conversion as the other pytorch/HF
consumers.

commit 68a1fa2e2eef0f0a027623be5f90c792884d6d21
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sat May 9 23:13:39 2026 +0200

dev-python/stop-sequencer: new package, evalplus dep