commit 947b4b351eaff207bb5a24df4ea5163bdd489afd
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Tue Jun 16 11:08:08 2026 +0200

dev-python/comfy-aimdo-bin: drop 0.4.9

commit 7b34ffdd57e89775cd3adfb9ed2a7a203d245d93
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Tue Jun 16 10:48:32 2026 +0200

dev-python/comfy-aimdo-bin: add 0.4.10

commit 0023bbbe8480d0dedcd453f57059b5474b15fbe0
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Mon Jun 15 14:14:51 2026 +0200

dev-python/comfy-aimdo-bin: add 0.4.9

commit 52776554669b16dc16876fe6583e248410f62b7e
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Mon Jun 15 00:16:02 2026 +0200

dev-python/comfy-aimdo-bin: add 0.4.8

PyTorch VRAM allocator with on-demand weight offloading. ComfyUI 0.24.x
hard-imports comfy_aimdo at module load (execution.py, model_management.py,
model_patcher.py, pinned_memory.py), so it is mandatory for a working install.

Upstream ships no sdist. USE=cuda installs the cp39-abi3 x86_64 manylinux CUDA
wheel (NVIDIA, needs a CUDA 12.8+ runtime, does the GPU VRAM offloading);
without cuda the py3-none-any pure-python wheel -- importable anywhere but a
no-op allocator -- for the CPU/ROCm fallback path.