dev-python/runai-model-streamer-bin
Run:ai's fast multi-tensor PyTorch model streamer (binary wheel)
Screenshots
View
Download
BrowseChangeLog
commit dfb96abfba96f17b9148f96ccdff4e2c40e7bc96
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Wed May 13 14:35:30 2026 +0200
dev-python/runai-model-streamer-bin: disable py3.11
commit b6e5192741d6c22890c721f1684d5065d122112c
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sun May 10 18:38:33 2026 +0200
dev-python/runai-model-streamer-bin: switch to DISTUTILS_SINGLE_IMPL
sci-ml/pytorch is SINGLE_IMPL; multi-impl consumer with bare
$ produces python_targets_python3_*(-)? that the
single-impl child can't expose. Replace python_foreach_impl install loop
with single-shot python_setup + $ -m installer (same shape as
the amd-quark-bin conversion in this round).
commit 7b47d63aa2674adca016da5208b5cf975f01d815
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Thu May 7 00:59:28 2026 +0200
dev-python/runai-model-streamer-bin: new package
Required by dev-python/vllm; upstream releases only manylinux2014_x86_64
wheels and a Bazel-based source build. Ships the wheel directly via
installer module.
Cloud backends ([s3,gcs,azure] extras) are deferred — vllm's local-
file model loading still works without them, and packaging
google-cloud-storage / azure-identity (with their large transitive
chains) is more cost than the missing backends are worth right now.
NonsolvableDepsInStable is policy noise — pytorch is ~amd64 only too.
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Wed May 13 14:35:30 2026 +0200
dev-python/runai-model-streamer-bin: disable py3.11
commit b6e5192741d6c22890c721f1684d5065d122112c
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sun May 10 18:38:33 2026 +0200
dev-python/runai-model-streamer-bin: switch to DISTUTILS_SINGLE_IMPL
sci-ml/pytorch is SINGLE_IMPL; multi-impl consumer with bare
$ produces python_targets_python3_*(-)? that the
single-impl child can't expose. Replace python_foreach_impl install loop
with single-shot python_setup + $ -m installer (same shape as
the amd-quark-bin conversion in this round).
commit 7b47d63aa2674adca016da5208b5cf975f01d815
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Thu May 7 00:59:28 2026 +0200
dev-python/runai-model-streamer-bin: new package
Required by dev-python/vllm; upstream releases only manylinux2014_x86_64
wheels and a Bazel-based source build. Ships the wheel directly via
installer module.
Cloud backends ([s3,gcs,azure] extras) are deferred — vllm's local-
file model loading still works without them, and packaging
google-cloud-storage / azure-identity (with their large transitive
chains) is more cost than the missing backends are worth right now.
NonsolvableDepsInStable is policy noise — pytorch is ~amd64 only too.