v0.1.7 — standalone load_pipeline fix
Fix: Standalone install actually works now
v0.1.6 shipped to the Comfy Registry but failed at first run with a misleading
Repository Not Found for url: https://huggingface.co/ckpts/... 404. v0.1.7
fixes the real root cause (three layers deep) and adds standalone-mode install support.
Root cause analysis
comfy-aimdo'smmgplib (shipped in ComfyUI Desktop's.venv)
monkey-patchessafetensors.torch.load_filewith a memory-mapped reader.
Its dtype map atmmgp/safetensors2.py:33omits complex dtypes; one
Pixal3D decoder checkpoint contains a singleC64(complex64) tensor →
KeyError: 'C64'blows up the load.- Upstream Pixal3D's
pipelines/base.py:43-46has a bareexcept Exception
that silently swallows thatKeyErrorand retries with the relative path. - The retry path in
models/__init__.py:60misparses relative paths like
"ckpts/ss_flow_..."as Hugging Face repo IDs and emits the bogus 404 we see.
Fixes (nodes/pixal3d_stages.py)
snapshot_download(model_path)resolves to an absolute local snapshot dir
beforefrom_pretrained, so the loader'sos.path.existscheck (relative
to CWD) always succeeds.- Pre-extend
mmgp.safetensors2._map_to_dtypewithC64/C128so the mmap
reader handles complex tensors. - Wrap
from_pretrainedintorch.inference_mode(False)as belt-and-suspenders
againstload_state_dictversion-counter errors on freshly-constructed
nn.Moduleparameters under ComfyUI's outerinference_mode.
Fixes (install.py)
- Standalone-mode fallback: if no
pozzettiandrea/ComfyUI-TRELLIS2
sibling exists, install deps intosys.executableinstead of hard-failing.
This matches how Comfy Registry / Manager invokeinstall.pywith the
.venv's python — fresh Registry installs now work out of the box. - Worker python search prefers
C:\ce\_env_*(the real comfy-env worker
pool) over any in-plugin stubScripts/python.exe. install_requirementssplitsrequirements.txtinto VCS / hashed / plain
buckets so pip's--require-hashesmode doesn't reject the VCS MoGe entry.- Error message:
visualbruno→pozzettiandrea(correct upstream fork).
Verified
End-to-end on ComfyUI Desktop + RTX 5090: 56.5s warm run with 8/8/8 steps +
1024 texture + low_vram=True → 8.6 MB GLB output.
Full Changelog: v0.1.6...v0.1.7