commit 50db0bf02ec15e5fd6e1fedfbf47e9b98888c608
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Thu Sep 10 04:46:31 2026 +0200

sci-ml/bigcode-eval: trim snapshot comment prose

commit a8dd60bdf3de1e2a59ef8565efe4aa5926214917
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Mon Sep 7 11:46:32 2026 +0200

sci-ml/bigcode-eval: Keyword 0_pre20250715 for ~arm64

commit 653ccd953c77cab7aaf1f6b6d3a22afa006f74d1
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Mon Sep 7 11:44:51 2026 +0200

sci-ml/bigcode-eval: set snapshot version metadata

commit d945ded628b9276d245e6f5da0b13bd3c0c4e3c2
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sun Jul 5 14:42:11 2026 +0200

sci-ml/bigcode-eval: dedup the pyext-drop rationale comment

commit 4594472bb37cb75509a517b7e30ea3255763c3c8
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sun May 10 14:51:11 2026 +0200

sci-ml/bigcode-eval: switch to DISTUTILS_SINGLE_IMPL

The full sci-ml stack here (accelerate, datasets, evaluate,
huggingface_hub, transformers, pytorch) is SINGLE_IMPL; depending on
them from a multi-impl ebuild yields python_targets_python3_*(-)? that
the children can't expose. Make bigcode-eval single-impl, put the entire
sci-ml block on bare $, and wrap the two
multi-impl helpers (mosestokenizer, fsspec) in python_gen_cond_dep.

commit 80345549e8aac49998ff392ab7767fa2f5d449a3
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sat May 9 22:15:27 2026 +0200

sci-ml/bigcode-eval: new package, BigCode evaluation-harness HEAD snapshot