sci-ml/trellis-cpp
TRELLIS.2 image to 3D in C++ and GGML
-
trellis-cpp-0.8.1~amd64cuda vulkan +webp weights-f16 weights-q4 weights-q8
View
Download
Browse License: MIT Overlay: tatsh-overlay
ChangeLog
commit 209c275b4821ec6eedcd24ed3077dfe087791e12
Author: Andrew Udvare <audvare@gmail.com>
Date: Mon Oct 12 14:36:30 2026 -0400
*/*: drop trailing period from DESCRIPTION
Signed-off-by: Andrew Udvare <audvare@gmail.com>
commit 1feff77e7f56beea86dbecd51c153693f4d6ca38
Author: Andrew Udvare <audvare@gmail.com>
Date: Mon Oct 12 04:28:28 2026 -0400
sci-ml/trellis-cpp: add 0.8.1, drop 0.6.0-r1
commit 9888bcfcb43e916190671c7bec8a19e5df376ef8
Author: Andrew Udvare <audvare@gmail.com>
Date: Sat Sep 19 18:52:15 2026 -0400
sci-ml/trellis-cpp: install the TRELLIS.2 weights
Fetch the GGUF weights from ilintar/trellis2-gguf, pinned to revision
a57397bd so the distfiles are immutable, and install them to
/usr/share/trellis-cpp/models. The model directory default that
trellis-cli and trellis-server share is relative ("models"), so patch it
to that path. Quantization is selected with the weights-f16, weights-q8
and weights-q4 USE flags; birefnet and ss_dec are identical across the
three and are fetched once.
Signed-off-by: Andrew Udvare <audvare@gmail.com>
commit 684cba1fe33e4ee1e3026740ae01720358f35a45
Author: Andrew Udvare <audvare@gmail.com>
Date: Wed Sep 16 10:22:51 2026 -0400
sci-ml/trellis-cpp: drop the version from patch filenames
PATCHES referenced $/$-..., which interpolates the current
version, so the next bump silently repoints it at files that do not exist and
only fails once src_prepare runs. Name the patches after $ instead.
Signed-off-by: Andrew Udvare <audvare@gmail.com>
commit cf86603ace003e241d2dbe5a8c281f547a13d5b5
Author: Andrew Udvare <audvare@gmail.com>
Date: Sun Sep 13 08:21:34 2026 -0400
*/*: fix shellcheck warnings
Signed-off-by: Andrew Udvare <audvare@gmail.com>
commit bbaef88c17f74aa593c40678a17017269c4b015a
Author: Andrew Udvare <audvare@gmail.com>
Date: Sun Sep 13 04:47:07 2026 -0400
sci-ml/trellis-cpp: new package, add 0.6.0
Signed-off-by: Andrew Udvare <audvare@gmail.com>
Author: Andrew Udvare <audvare@gmail.com>
Date: Mon Oct 12 14:36:30 2026 -0400
*/*: drop trailing period from DESCRIPTION
Signed-off-by: Andrew Udvare <audvare@gmail.com>
commit 1feff77e7f56beea86dbecd51c153693f4d6ca38
Author: Andrew Udvare <audvare@gmail.com>
Date: Mon Oct 12 04:28:28 2026 -0400
sci-ml/trellis-cpp: add 0.8.1, drop 0.6.0-r1
commit 9888bcfcb43e916190671c7bec8a19e5df376ef8
Author: Andrew Udvare <audvare@gmail.com>
Date: Sat Sep 19 18:52:15 2026 -0400
sci-ml/trellis-cpp: install the TRELLIS.2 weights
Fetch the GGUF weights from ilintar/trellis2-gguf, pinned to revision
a57397bd so the distfiles are immutable, and install them to
/usr/share/trellis-cpp/models. The model directory default that
trellis-cli and trellis-server share is relative ("models"), so patch it
to that path. Quantization is selected with the weights-f16, weights-q8
and weights-q4 USE flags; birefnet and ss_dec are identical across the
three and are fetched once.
Signed-off-by: Andrew Udvare <audvare@gmail.com>
commit 684cba1fe33e4ee1e3026740ae01720358f35a45
Author: Andrew Udvare <audvare@gmail.com>
Date: Wed Sep 16 10:22:51 2026 -0400
sci-ml/trellis-cpp: drop the version from patch filenames
PATCHES referenced $/$-..., which interpolates the current
version, so the next bump silently repoints it at files that do not exist and
only fails once src_prepare runs. Name the patches after $ instead.
Signed-off-by: Andrew Udvare <audvare@gmail.com>
commit cf86603ace003e241d2dbe5a8c281f547a13d5b5
Author: Andrew Udvare <audvare@gmail.com>
Date: Sun Sep 13 08:21:34 2026 -0400
*/*: fix shellcheck warnings
Signed-off-by: Andrew Udvare <audvare@gmail.com>
commit bbaef88c17f74aa593c40678a17017269c4b015a
Author: Andrew Udvare <audvare@gmail.com>
Date: Sun Sep 13 04:47:07 2026 -0400
sci-ml/trellis-cpp: new package, add 0.6.0
Signed-off-by: Andrew Udvare <audvare@gmail.com>

