aszc
02b6e8ece3
Modernizes ComfyUI-CoreMLSuite onto Python 3.12 / torch 2.7 / coremltools 9 with a characterization-test safety net. The default conversion path is unchanged; existing saved workflows produce identical output. - Toolchain bump (Python 3.12, torch 2.7, coremltools 9, numpy <2) with the blocking upstream pins overridden. - Framework-free logic moved into coreml_suite/core/ (no comfy/coremltools imports); old module paths re-export from there. - Opt-in quantize_nbits dropdown (none|8|6|4) for k-means weight palettization; default none is byte-for-byte identical to before. - Tiered CI: Tier 0 (Linux unit), Tier 1 (macOS-ARM smoke), Tier 2 (self-hosted Apple Silicon golden-image check on the ANE).
19 lines
847 B
Plaintext
19 lines
847 B
Plaintext
# Toolchain ceiling for installing a floating-latest ComfyUI's requirements.txt
|
|
# in the Tier 2 nightly canary (.github/workflows/tier2.yml, latest mode).
|
|
#
|
|
# ComfyUI's requirements.txt requests bare `torch`/`torchvision`/`torchaudio`
|
|
# and `numpy>=1.25.0`, which would float past the versions coremltools 9 /
|
|
# apple-ml-stable-diffusion have been validated against.
|
|
# These constraints cap the resolution so the canary keeps testing the same
|
|
# toolchain the suite actually ships.
|
|
#
|
|
# If upstream ComfyUI ever hard-requires something beyond these bounds, the
|
|
# install FAILS — and that failure is the signal we want: it means the host
|
|
# outgrew the pinned toolchain and coremltools / ml-stable-diffusion need a
|
|
# deliberate bump, not a silent float.
|
|
torch>=2.7,<2.8
|
|
torchvision>=0.22,<0.23
|
|
torchaudio>=2.7,<2.8
|
|
numpy>=1.25,<2
|
|
coremltools>=9,<10
|